<<

University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2020-01 A multimethod analysis to assess locomotor capabilities in stem from Blue Beach (Tournaisian; Early ), Nova Scotia

Lennie, Kendra Ilana

Lennie, K. I. (2020). A multimethod analysis to assess locomotor capabilities in stem tetrapods from Blue Beach (Tournaisian; Early Carboniferous), Nova Scotia (Unpublished master's thesis). University of Calgary, Calgary, AB. http://hdl.handle.net/1880/111587 master thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

A multimethod analysis to assess locomotor capabilities in stem tetrapods from Blue Beach

(Tournaisian; Early Carboniferous), Nova Scotia

by

Kendra Ilana Lennie

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF SCIENCE

GRADUATE PROGRAM IN BIOLOGICAL SCIENCES

CALGARY, ALBERTA

JANUARY, 2020

© Kendra Ilana Lennie 2020 Abstract In vertebrate evolution the fin-to-limb transition was an important precursor to the diversification and radiation of terrestrial animals into novel environments. This transition began in the Devonian and continued through the Carboniferous and involved physiological and biomechanical changes. I used a multi-method approach to assess external and internal limb bone features to evaluate Early Carboniferous (Tournaisian) limb bones from Blue Beach and associated them with aquatic to terrestrial lifestyles. Tournaisian material was collected at Blue Beach located near Hantsport, Nova Scotia, but much of it has not been formally described because the disarticulated and isolated tetrapod elements made identification to the species level difficult. In this thesis I described new morphotypes attributable to the family level which are used in the following chapters. Once the external morphology of the Blue Beach bones was described I compared them with the femora of extant aquatic, amphibious, and terrestrial tetrapods to evaluate which locomotor behaviour the fossil femora most resembled. I additionally examined cross-sectional bone profiles of Blue Beach tetrapod femora to infer lifestyle. Midshaft analyses relied on a single two-dimensional image to represent a dynamically structured bone so

I also used a novel method for assessing three-dimensional trabecular data to qualitatively and quantitatively infer lifestyle from the Blue Beach femora. From the various analyses of internal and external bone morphology it was clear that external bone features of modern and early fossil tetrapod femora are dissimilar, which lead to difficulties in drawing conclusions based off external qualitative data. Internal data, from two-dimensional midshaft and three-dimensional trabecular structures, produced quantitative results that lead to the same conclusion, that the Blue

Beach femora are consistent with those of aquatic animals. This implies that the initial diversification of the tetrapod body plans present in the Early Carboniferous was not the result of terrestrialization but appears to have preceded it.

ii

Preface This thesis is original, unpublished, work by the author, K. Lennie.

iii

Acknowledgments I would like to thank my supervisor Jason Anderson for helping me revise countless drafts and being patient with me for the last two and a half years. I would also like to thank my committee members Jessica Theodor, Sarah Manske, and Heather Jamniczky, whose support and suggestions (especially about trabecular bone analysis) was critical to the success of my project.

I would like to thank NSERC for funding a Discovery Grant to Jason Anderson, which made this work possible. Additionally I’d like to thank Natalie Reznikov for training on the Bone Analysis module of Dragonfly ORS (Montreal), Katherine Ogdon and Tim Fedak for access to the specimens at the Nova Scotia Museum, Chris Mansky and Sonja Wood for access and preparation work on the specimens from the Blue Beach Fossil Museum, and Johanne Kerr for help finding and accessing elements from Eusthenopteron at Musée d’Histoire Naturelle de

Miguasha. I want to thank my lab mates Jason Pardo, Conrad Wilson, James Campbell, Ramon

Nagesan, Amber Whitbone, and Jamey Creighton, and Keith Johnstone for many informative and thought-provoking discussions, for listening to my constant complaining about muscles, and for providing feedback on many written applications and assignments. I would finally especially like to thank my parents for pushing me to do my best and supporting me in all my endeavours.

iv

Dedication I dedicate this thesis to all the girls and women who are defying traditional gender roles in science and sport. Never let anyone tell you should not or cannot do something because you are a girl.

v

Table of Contents Abstract ...... ii Preface ...... iii Acknowledgments...... iv Dedication ...... v Table of Contents ...... vi List of Tables ...... ix List of Figures ...... x Table of extant animals ...... xiv Abbreviations ...... xv Institutional ...... xv Anatomical ...... xv Variables ...... xv Chapter 1: Introduction ...... 1 Introduction ...... 1 Historic hypotheses ...... 2 Romer’s Gap ...... 2 Blue Beach ...... 3 Historical proxies for locomotion ...... 6 Predictions ...... 11 Literature cited ...... 13 Chapter 2: Descriptions of new tetrapod limb elements from Blue Beach, Nova Scotia ...... 23 Introduction ...... 23 Abbreviations ...... 23 Geological context ...... 24 Material and Methods ...... 25 Results ...... 25 Humeri ...... 25 Ulna ...... 26 Radii ...... 27 Femora ...... 29 Fibulae ...... 41 Tibiae ...... 44 Discussion ...... 48

vi

Conclusion ...... 50 Literature Cited ...... 51 Chapter 3: Musculature comparisons using extant taxa ...... 56 Introduction ...... 56 Materials and Methods ...... 59 Results ...... 62 Large Osteological Features ...... 63 Discussion ...... 68 Gross patterns in bone morphology in representative extant taxa and comparison to the Blue Beach material...... 68 Extensive pitting of early tetrapod femora ...... 72 Conclusions ...... 73 Literature Cited ...... 75 Chapter 4: Midshaft profiles ...... 82 Introduction ...... 82 Abbreviations ...... 86 Materials and Methods ...... 87 Results ...... 93 NSM.005.GF.045.044 ...... 94 NSM.005.GF.045.047-Distal ...... 94 NSM.007.GF.004.630 ...... 95 NSM.005.GF.045.048-A ...... 96 RM.20.6711 ...... 96 YPM.PU.20103 ...... 96 Discussion ...... 98 Conclusion ...... 100 Literature Cited ...... 101 Chapter 5: Trabecular Analysis ...... 109 Introduction ...... 109 Wolff’s law ...... 112 Anisotropy ...... 112 Fin-to-limb ...... 114 Abbreviations ...... 115 Locality ...... 116 Materials and Methods ...... 116

vii

Results ...... 126 Extant terrestrial taxa ...... 126 Extant aquatic taxa ...... 130 Fossil bones ...... 134 Principal Component Analysis ...... 143 Discussion ...... 147 Femora ...... 147 Forelimb elements ...... 152 General observations ...... 153 Principal Component Analysis ...... 153 Conclusions ...... 156 Literature cited ...... 157 Chapter 6: Summary and Conclusions...... 170 Chapter 1 summary ...... 170 Chapter 2 summary ...... 170 Chapter 3 summary ...... 171 Chapter 4 summary ...... 172 Chapter 5 summary ...... 174 Concluding Remarks ...... 176 Literature Cited ...... 177 Literature Cited ...... 180 Appendix A ...... 195 Appendix B ...... 197 Appendix C ...... 229 Appendix D ...... 242 Appendix E ...... 266 Appendix F ...... 283 Appendix G ...... 288

viii

List of Tables

Table 2.1. Borrowed Blue Beach elements and their attributed taxonomic morphologies and lifestyles. . 47 Table 4.1 Femora used for midshaft profiles, and their historical locomotor behaviour associations...... 89 Table 4.2. ImageJ threshold functions and values of midshaft slices...... 91 Table 4.3. Angular Compactness Profiles of Blue Beach Femora Midshafts ...... 95 Table 5.1. Variables of whole bone analysis ...... 124 Table 5.2. Summary of Primary Component Analyses ...... 144 Table 5.3. PCA loadings ...... 145 Table S.0.1 PCA summary statistics ...... 283 Table S.0.2 PCA loadings ...... 283 Table S.1 PCA no Felis summary ...... 288 Table S.2 PCA no Felis loadings ...... 288

ix

List of Figures Figure 1.1. The outgroups and sister group to early tetrapods that are required to infer soft tissues using the extant phylogenetic bracket...... 10 Figure 2.1. Micro computed tomography (µCT) images of NSM.007.GF.004.733: right ulna. Tulerpeton- like ulna in (A) extensor, (B) anterior, (C) flexor, (D) posterior, (E) proximal and (F) distal view. mr, mesial ridge. Scalebar 10mm...... 27 Figure 2.2. µCT images of NSM.014.GF.036.02, right Acanthostega-like radius (A-F) and NSM.007.GF.004.767, left radius (G-L) in (A, G) extensor, (B, H) posterior, (C, I) flexor, (E, K) proximal and (F, L) distal view. vrc, ventral radial crest. Scalebar 10mm...... 28 Figure 2.3. µCT images of NSM.005.GF.045.048-A, left embolomerous femur (A-F), and RM.20.5222, a right Acanthostega-like) femur (G-L) in (A, G) extensor, (B, H) anterior, (C, I) flexor, (D, J) posterior, (E, K) proximal, and (F, L) distal views. ab, adductor blade; ac, adductor crest; icf, intercondylar fossa; itf, intertrochanteric fossa; pa, popliteal area. Scalebar 10mm...... 30 Figure 2.4. µCT images of NSM.007.GF.004.630 a right Tulerpeton-like femur in (A) extensor, (B) anterior, (C) flexor, (D) posterior, (E) proximal, and (F) distal view. ab, adductor blade; ac, adductor crest; cp, central prominence; ff, fibular facet; fo, fibular fossa; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm...... 32 Figure 2.5. µCT images of YPM.PU.20103 a right Tulerpeton-like/whatcheeriid femur (A-F) and NSM.005.GF.045.042 a left Tulerpeton-like femur (G-L) in (A,G) extensor, (B,H) anterior, (C,I) flexor, (D,J) posterior, (E,K) proximal, and (F,L) distal views. scalebar 10mm...... 34 Figure 2.6. µCT images (meshes) of YPM.PU.20103 (A-D) and NSM.005.GF.045.042 (E-H) in (A,E) extensor, (B,F) anterior, (C,G) flexor and, (D,H) posterior views. ab, adductor blade; ac, adductor crest; der, dorsal elongate ridge; ff, fibular facet; fo, fibular fossa; icf, intercondylar fossa; it, internal trochanter; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm...... 36 Figure 2.7. µCT images of NSM.005.GF.045.044, distal end of a left Ossinodus-like femur (A-E), and NSM.005.GF.045.047 proximal (K-O) and distal (F-J) segments in (A,F,K) extensor, (B,G,L) anterior, (C,H,M) flexor, (D,I,N) posterior, (E,J) distal and (O) proximal views. ab, adductor blade; ac, adductor crest; cp, central prominence; ff, fibular facet; fo, fibular fossa; ft, fourth trochanter; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm...... 39 Figure 2.8. µCT images of NSM.005.GF.045.061 left Pederpes-like fibula (A-F) in and left fibula NSM.005.GF.045.063 (G-L) in (A,G) extensor, (B,H) posterior, (C,I) flexor, (D,J) anterior, (E,K) proximal, and (F,L) distal views. Scalebar 10mm...... 43 Figure 2.9. µCT images of left tibiae RM.20.2555 (A-F) and NSM.005.GF.045.320 (G-L) in (A,G) extensor, (B,H) anterior, (C,I) flexor, (D,J) posterior, (E,K) proximal and (F,L) distal views. Scalebar 10mm ...... 45 Figure 3.1 Cladogram showing nested groups more closely related to each other than the taxa to the left. For example, birds and crocodiles form a sister group to lepidosaurs, and are more closely related to each other than they are to lepidosaurs. Lepidosaurs, crocodiles, and birds form a sister groups with mammals, and are more closely related to each other than they are to mammals. And so on. The silhouettes illustrate there is a great deal of variation in body plans, and all major extant groups have some taxa that have undergone a secondary transition to an aquatic or amphibious lifestyle...... 57 Figure 3.2. Progression of sampled aquatic to amphibious to terrestrial femora sampled. µCT images ofA, femur of Necturus (aquatic). B, femur of Cynops (amphibious). C, femur of Uromastyx (terrestrial). D, femur of Eublepharis (terrestrial). Scale bars represent 1mm...... 63

x

Figure 3.3. Comparison of topological similarities between the femora of Cynops (a,c) and Blue Beach YPM.PU.20103 (b,d). Yellow appears to resemble an intertrochanteric fossa, blue the popliteal area, and purple a fibular fossa. Red represents well developed ridges ...... 66 Figure 3.4. Tetrapod femur NSM.007.GF.004.630. A, green arrows indicate where large rugose pits extend through the cortical bone. B, same pits in A visible in transaxial sections crossing through the cortical layer. C, same pits viewed in longitudinal sections crossing the cortical layer...... 69 Figure 3.5. Femur NSM.007.GF.004.630. Green arrows are pointing to small “pits” found across blue beach taxa, similar in size to those described by Bishop (2014). B, same pits in A visible in transaxial sections crossing through the cortical layer. C, same pits viewed in longitudinal sections crossing the cortical layer...... 71 Figure 4.1 Bone Profiler analysis. A, Schematic of the 51 ring and 6-degree increments analysed by bone profiler. B, Graphical representation of the four compactness parameters. Modified from Quemeneur et al. (2013)...... 88 Figure 4.2. Example of how a midshaft (RM.20.6711) was analysed in bone profiler. A, original midshaft image. B, midshaft after contrast was adjusted. C, binary midshaft after threshold applied. D, binary image uploaded into Bone Profiler and bone parameters automatically generated. E, global and angular output from Bone Profiler...... 93 Figure 4.3. Bone Profiler graphs illustrating the compactness profiles of the Blue Beach fossil Femora, with a platypus (MVZ:32885) included for comparison representing an aquatic bone profile, and cat representing terrestrial...... 94 Figure 4.4. Unmodified µCT (A,C,E,G, I, and K) and thresholded binary images (B,D,F,H, J, and L) of the midshafts of fossil femora...... 97 Figure 5.1. Coronal slice of the proximal femur of a juvenile cat. Highlighted in red is highly aligned trabecular structure that is indicative of the force transmitted from pelvis to femur in response to walking in a terrestrial environment (i.e., under gravitational forces)...... 111 Figure 5.2. Transaxial slice of fossil tibia NSM.014.GF.036.004. trabecular structure is highlighted in pink. A shows that trabecular structure is left with aligned edges where it is cut from the trabecular structure in segmentation. B shows how this effects anisotropy calculations at the edges where these cuts get analysed and produce areas of high anisotropy. C shows the inverted trabecular area (grey-blue) that will be overwritten on the anisotropy heat map. D shows the more accurate overwritten anisotropy heat map...... 123 Figure 5.3. Anisotropy vector maps and histograms of extant terrestrial humeri. A-D Eublepharis humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis humerus. F-I, Uromastyx humerus. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx humerusFigure 5.2. Transaxial slice of fossil tibia NSM.014.GF.036.004. trabecular structure is highlighted in pink. A shows that trabecular structure is left with aligned edges where it is cut from the trabecular structure in segmentation. B shows how this effects anisotropy calculations at the edges where these cuts get analysed and produce areas of high anisotropy. C shows the inverted trabecular area (grey-blue) that will be overwritten on the anisotropy heat map. D shows the more accurate overwritten anisotropy heat map. . 123 Figure 5.3. Anisotropy vector maps and histograms of extant terrestrial humeri. A-D Eublepharis humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis humerus. F-I, Uromastyx humerus. G, vector map of all anisotropy. H, Vector

xi map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx humerus...... 127 Figure 5.4. Anisotropy vector maps and histograms of extant terretrial femora. A-D Eublepharis femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis femur. F-I, Uromastyx femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx femur. K-N, Felis femur. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Felis femur...... 129 Figure 5.5. Anisotropy vector maps and histograms of extant aquatic humeri. A-D Cryptobranchus humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Cryptobranchus humerus. F-I, Ornithorhynchus humerus. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Ornithorhynchus humerus. K-N, Amblyrhynchus humerus. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Amblyrhynchus humerus...... 131 Figure 5.6. Anisotropy vector maps and histograms of extant aquatic femora. A-D Cryptobranchus femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Cryptobranchus femur. F-I, Ornithorhynchus femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Ornithorhynchus femur. K-N, Amblyrhynchus femur. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Amblyrhynchus femur...... 133 Figure 5.7. Anisotropy vector maps and histogram of a Eusthenopteron humerus. B, vector map of all anisotropy. C, vector map of anisotropy 0.65 and above in the flexor half of the bone. D, vector map of anisotropy 0.65 and above in the extensor half of the bone. E, anisotropy frequency histogram of Eusthenopteron humerus...... 134 Figure 5.8 Anisotropy vector maps and histograms of fossil forelimb elements. A-D, YPM.PU.23545 whatcheerid humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of YPM.PU.23545. F-I, NSM.007.GF.004.733 Tulerpeton-like ulna. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.007.GF.004.733. K-N, NSM.014.GF.036.002 Acanthostega-like radius. L, vector map of all anisotropy. M, vector map of anisotropy 0.65 and above in the flexor half of the bone. N, vector map of anisotropy 0.65 and above in the extensor half of the bone. O, Anisotropy frequency histogram of NSM.014.GF.036.002...... 135 Figure 5.9. Anisotropy vector maps and histograms of fossil femora. A-D, NSM.005.GF.045.048-A embolomerous femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.048-A. F-I, NSM.007.GF.004.630 indet femur. G,

xii vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.007.GF.004.630...... 137 Figure 5.10. Anisotropy vector maps and historgrams of partial fossil femora. A-D, NSM.005.GF.045.044 Ossinodus-like distal femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.044. F-I, NSM.005.GF.045.047 whatcheerid distal femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.005.GF.045.047. K-N, NSM.005.GF.045.047 whatcheerid proximal femur. L, vector map of all anisotropy. M, vector map of anisotropy 0.65 and above in the flexor half of the bone anisotropy 0.65 and above in the flexor half of the bone. N, vector map of anisotropy 0.65 and above in the extensor half of the bone. O, Anisotropy frequency histogram of NSM.005.GF.045.047...... 139 Figure 5.11. Anisotropy vector maps and histograms of fossil fibulae. A-D, NSM.005.GF.045.112 Crassigyrinus-like fibula. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.112. F-I, NSM.005.GF.045.061 Pederpes-like fibula. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.005.GF.045.061...... 141 Figure 5.12. Anisotropy vector maps and histograms of fossil tibiae A-D, RM.20.5222 Proterogyrinus- like tibia. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of RM.20.5222. F-I, NSM.014.GF.036.004 “type 2” whatcheerid tibia. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.014.GF.036.004...... 142 Figure 5.13. PCA plots of all limb elements used. Brown corresponds to terrestrial animals, green to amphibious, blue to aquatic and purple to fossil tetrapod elements. Summary and loadings in table 1 and 2...... 144 Figure 5.14. PCA plots of femora only. Brown corresponds to terrestrial animals, green to amphibious, blue to aquatic and purple to fossil tetrapod elements. Summary and loadings in table 1 and 2...... 146 Figure 5.15. Platypus femur. A, Muscle map adapted from Gambaryan et al (2002). Only the muscles corresponding to the aligned anisotropy vectors are illustrated here. B, vector map of the high anisotropy vectors (0.65 and above) of the whole femur...... 150

xiii

Table of extant animals Genus Common name Felis House Cat Uromastyx Spiny-Tailed Lizard Eublepharis Leopard Gecko Amblyrhynchus Marine Iguana Ornithorhynchus Platypus Lutra River Otter Australian Snapping Elseya Turtle

xiv

Abbreviations

Institutional

Blue Beach Fossil Museum, Hantsport, Nova Scotia, Canada (BBFM); Musée d’Histoire

Naturelle de Miguasha, Québec, Canada (MHNM); Museum of Vertebrate Zoology, University of California, Berkeley, USA (MVZ); Nova Scotia Museum of Natural History, Halifax, Canada

(NSM); Redpath Museum, McGill University, Montreal, Canada (RM); University of Michigan

Museum of Zoology, Ann Arbor, Michigan, USA (UMMZ); Yale Peabody Museum of Natural

History, New Haven, Connecticut, USA (YPM); Museum für Naturkunde Berlin, Germany

(ZMB).

Anatomical

ab, adductor blade; ac, adductor crest; cn, cnemial crest; cp, central prominence; der, dorsal elongate crest; ff, fibular facet; fo, fibular fossa; ft, fourth trochanter; icf, intercondylar fossa; it, internal trochanter; itf, intertrochanteric fossa; mr, mesial ridge; pa, popliteal area;

PIFE, Puboischiofemoralis externus; PIFI, Puboischiofemoralis internus; tf, tibial facet.

Variables

Ani.M, mode of frequency of anisotropy; BV, bone volume; BV/TV, bone volume fraction; Ct.Ar/Tt.Ar, cortical area fraction; Max, global maximum asymptote; Min, global minimum asymptote; P, global transition; Rmax, angular maximum asymptote; Rmin, angular minimum asymptote; RP, angular transition point; RS, angular 1/slope; point; S, global 1/slope;

TV, total volume.

xv

Chapter 1: Introduction Introduction

The Devonian-Carboniferous transition occurred 360 million years ago. Sarcopterygian

(lobe-finned) fish arose in the Devonian period. This group includes the panderichthyids, a group with proximal bony elements in their fins proposed to be homologous to modern tetrapod elements (Boisvert et al., 2008). By the end of the Devonian the first digited tetrapods

Acanthostega, , and Tulerpeton were present in aquatic environments. The subsequent animals from the Carboniferous Era ranged from aquatic to terrestrial, but exactly how the transition to land took place remained a mystery because of a prominent gap in the fossil record. This gap is now being populated by fossils, but until recently paleontologists were unable to reliably infer locomotor behaviour of early fossil tetrapods from proxies other than external morphological features (Germain and Laurin, 2005; Kriloff et al., 2008; Laurin et al., 2011;

Sanchez et al., 2012, 2013; Quemeneur et al., 2013; Plasse et al., 2019). Analyses of internal structure of limb bones, such as midshaft analyses, show that the internal structure of limb bones provides a robust way to predict locomotor behaviour of tetrapods. Here I expand on that by using three-dimensional data from the internal structures of tetrapod limb bones from Blue

Beach, Nova Scotia, to infer locomotor potential and behaviour in early tetrapods.

In this thesis a “limb” is an appendage consisting of a stylopod, zeugopod, and most importantly a digited autopod. Locomotor behaviour refers to habitual locomotion in a general

(aquatic, amphibious, or terrestrial) environment, and tetrapod is used to describe any animal with four limbs.

1

Historic hypotheses

Historically the fin-to-limb and water-to-land transitions were thought to be contemporaneous events. The “drying pools hypothesis” for example, imagined sarcopterygian fish traversing small areas of land in order to get from small pools to bigger pools for supporting life (Inger, 1957; Barrell et al., 2011). The idea was that lobe finned fish with more limb-like fins successfully crossed the terrestrial landscape and were able to reproduce. Under the drying pools hypothesis any animal with digited limbs should be terrestrial, or at least amphibious. The discovery of internal gills in Acanthostega (a digited tetrapod) forced the re-evaluation of this traditional idea (Coates and Clack, 1991).

Romer’s Gap

Research into the fin-to-limb transition is complicated by a paucity of material in the fossil record during the time when it was thought that the transition from water to land took place. Romer’s Gap, a phenomenon named for the man who observed it, Alfred Sherwood

Romer (Romer, 1956; Coates and Clack, 1995), originally ranged from the end of the Devonian to the Late Carboniferous. For Romer, Ichthyostega and Acanthostega represented Devonian aquatic animals from the beginning of the fin-to-limb transition and Seymouria, Archeria, and

Eryops represented later, more amphibious body forms from the Carboniferous to Permian

(Romer, 1956). Romer noted the existence of poorly preserved embolomere material in the

Carboniferous but doubted the accuracy of reconstructions at the time of his publication in 1956.

This left most of the Carboniferous period deficient in fossil material. Mid to late Carboniferous tetrapods slowly filled the latter portion of Romer’s Gap in the second half of the twentieth century (Holmes, 1984; Panchen, 1985; Romer and Godfrey, 1989; Panchen and Smithson,

1990; Lombard and Bolt, 1995); however, known isolated elements remained unstudied (see

2

remarks on Blue Beach material in Lebedev and Coates 1995; Clack and Carroll 2000; Warren and Turner 2004; Coates et al., 2008) until descriptions (Smithson and Clack, 2013; Anderson et al., 2015) and morphotype association by Anderson et al. (2015).

In the early 2000s Romer’s Gap was rapidly filled on both sides by a diverse assemblage of tetrapod material (Clack, 2001, 2002, 2006; Clack and Ahlberg, 2004; Warren and Turner,

2004; Clack and Finney, 2005; Smithson and Clack, 2013, 2018; Anderson et al., 2015; Clack et al., 2018). These fossils refuted other proposed explanations of Romer’s Gap, including a period of low oxygen (Ward et al., 2006), and an extended post extinction recovery period (Sallan et al.,

2010) as well as additional evidence, such as charcoal from wildfires, which require high levels of oxygen in the atmosphere (Mansky and Lucas, 2013). The currently emerging view is that a collection bias towards complete articulated material was responsible for Romer’s Gap

(Smithson et al., 2012).

Blue Beach

Blue Beach, Nova Scotia, is currently the oldest known locality within Romer’s Gap. It is early-mid to mid Tournaisian in age, extending further back than the CM (claviger - macra) biostratigraphical palynozone of similar localities in Britain (Utting et al., 1989; Martel and

Gibling, 1996). However, the Ballagan Formation of Scotland may include some of the slightly older PC (pretiosus - clavata) palynozone (Waters et al., 2009 ongoing). The Ballagan Formation

(which includes a number of sites; see Clack et al., 2017 supplementary information) yields articulated early tetrapod material from the Tournaisian. The dating of the Scottish formations is currently under review (TW:eed Project) and regions of the Ballagan Formation may actually be older than Blue Beach (Waters et al., 2009; Bennett et al., 2016; Clack et al., 2017).

3

Blue Beach is located along an estuary off the Bay of Fundy near Windsor, Nova Scotia.

Heavy tidal action at the site frequently exposes new fossil material (Mansky and Lucas, 2013).

At the time of deposition Blue Beach was a marginally marine environment, ranging from on- shore aerially exposed, to off-shore facies described as lagoonal (Tibert and Scott, 1999;

Mossman and Grantham, 2008), and often stormy (Mansky and Lucas, 2013). These conditions explain the presence of aquatic, amphibious, and possibly terrestrial vertebrates at the same locality. Animals living (or that died) in a nearshore but terrestrial environment could be washed away from nearby locations and deposited at Blue Beach. Aquatic animals washed ashore in previous storms, which had time to decompose between events could subsequently be deposited at Blue Beach as well. Trackways suggest the presence of terrestrial taxa (Mossman and

Grantham, 2008; Mansky and Lucas, 2013). The high energy depositional setting in conjunction with the disarticulated but minimally worn material suggest the animals decomposed before they were transported and deposited at Blue Beach (Mansky and Lucas, 2013). Limb bones such as humeri and femora are disarticulated but identifiable approximately to ordinal level and by morphotypes described in Anderson et al (2015). These identifications will be applied to the material used in this study.

Blue Beach preserves several elements that are similar to Devonian taxa Acanthostega and Tulerpeton (Anderson et al., 2015), which have been previously described as aquatic based on the presence of internal gills, and the depositional setting of the material (Coates and Clack,

1991; Lebedev and Coates, 1995). Blue Beach also has a diverse array of animals similar to some of the material known from the early Carboniferous, presumed to be more terrestrial. There appears to be a community of whatcheeriid-like animals that includes both aquatic (Whatcheeria;

Lombard and Bolt, 1995) and more terrestrial (Pederpes; Clack and Finney, 2005) tetrapods.

4

Other Carboniferous morphotypes include elements comparable to more amphibious Eoherpeton

(Smithson, 1985) and Proterogyrinus (Holmes, 1984).

Such a range of morphotypes and expected locomotor behaviours makes Blue Beach an ideal locality for investigations into limb bone structure and morphology associated with the evolution of terrestrial locomotion during the fin-to-limb transition. Additionally, there is a known footprint record at Blue Beach, suggesting the likelihood of finding fossils of at least partially terrestrial animals (Mossman and Grantham 2008; Mansky and Lucas 2013). The potential diversity of terrestrial animals will have implications for interpretations of the sequence of events critical to transition onto land. For example, the Blue Beach animals could provide clarification if a single group of animals transitioned onto land and then radiated, or if a diverse group of animals appear to contemporaneously transition to a terrestrial lifestyle. Additionally, the site controls for time so the animals present at Blue Beach represent diverse individuals coexisting at one time potentially across a range of environments, rather than changing fauna or ancestor-descendant relationships (i.e., they do not represent a single lineage changing over time).

In addition to a range of diverse morphotypes identifiable by external morphology, the

Blue Beach tetrapod elements are excellently preserved in three-dimensions. This is critical for the analyses used in chapters four and five, as internal morphology preserves more robust signals of locomotor behaviour than external morphology in many fossil elements (Zeininger et al.,

2016; Su and Carlson, 2017; Kivell et al., 2018; Dunmore et al., 2019; Georgiou et al., 2019)

5

Historical proxies for locomotion

To transition from an aquatic locomotor behaviour to a terrestrial one, the orientation of fins and their musculature must change. The fins of aquatic fish-like tetrapods must modify from a lateral paddling movement through a viscous medium, to a ventrally-oriented, crawling movement on land. The change in limb orientation from lateral to ventrolateral should be reflected in a change in direction of muscle forces acting on the limbs. Limbs used for locomotion on land experience the force of gravity in a more unidirectional manner as opposed to hydrodynamic forces acting equally in all directions. Additionally, gravity and ground reaction forces should result in forces concentrated on the proximal and distal ends of limb bones, indicating gravitational forces acting across joints.

The main reason proxies are used to infer physiological or biomechanical information such as locomotor behaviour is that soft tissues involved in such processes are not preserved.

Organic material of the bone breaks down and is replaced with mineral precipitates, except in rare instances (e.g. laggerstatten). The primary historical proxy for locomotor behaviour was the presence of digits. The presence of internal gills in addition to a digited limb in Acanthostega forced the need to investigate other proxies for locomotor behaviour. Subsequently proposed proxies include: digit number, limb proportion, degree of ossification, and proportion of compact bone to trabecular bone (Coates and Clack, 1990; Laurin et al., 2004, 2011; Warren and Turner,

2004; Clack and Finney, 2005; Quemeneur et al., 2013).

In Devonian tetrapods the number of digits on a limb was variable (Coates and Clack,

1990). Devonian stem tetrapods Acanthostega, Ichthyostega, and Tulerpeton, were polydactylous, with more than five (up to eight in Acanthostega; Clack, 2006) digits. Coates and

Clack (1990) concluded that pentadactyly (limbs with five digits) was not the primitive state for

6

early tetrapods, but it was later in the Carboniferous that pentadactyly became the predominant state for tetrapods (Coates and Clack, 1990). This pattern of dactyly was proposed as a proxy for lifestyle inference (Coates and Clack, 1990). More digits (polydactyly) could be used to infer aquatic animals because increased surface area was better for paddling limbs (Smithson et al.,

2012). However, it is unclear if pentadactylous limbs are a necessary precursor to terrestriality, or if terrestrial tetrapods share the pentadactylous condition because selection acted on it after tetrapods transitioned onto land (i.e., plesiomorphy or parallelism).

Limb bone ratios are another proxy used to indicate potential for terrestrial locomotion

(Clack, 2001; Warren and Turner, 2004; Warren, 2007; Sanchez et al., 2010). Shorter proximal elements are associated with aquatic tetrapods, whereas shorter distal elements relative to the proximal element are associated with terrestriality (Clack, 2001; Warren and Turner, 2004;

Warren, 2007; Sanchez et al., 2010). However, conclusions drawn from limb bone proportions may contradict those drawn from dactyly, robusticity, degree of ossification, and histological analysis (Sanchez et al., 2010).

Another proxy for assessing locomotor capabilities is robusticity (Lombard and Bolt,

1995; Milner and Lindsay, 1998; Clack, 2002; Clack and Finney, 2005; Warren, 2007; Shubin et al., 2014). Robusticity is poorly defined, qualitative, and subject to interpretation. In many cases it can be synonymized with well preserved or thick (Clack, 2001, 2002; Daeschler et al., 2006;

Pierce et al., 2012; Shubin et al., 2014). Robusticity is always used in conjunction with another proxy.

The degree of ossification of a tetrapod limb bone is also used to infer lifestyle (Laurin et al., 2004; Coates et al., 2008; Sanchez et al., 2010). Weakly ossified limb bones are associated with aquatic lifestyles, whereas more heavily ossified bones are associated with terrestrial

7

locomotion because they can withstand greater strain without breaking (Laurin et al., 2004;

Sanchez et al., 2010). Ontogeny may confound these assessments because juveniles have smaller and incompletely ossified limbs regardless of locomotor behaviour (Long and Gordon, 2004;

Warren, 2007).

In Early Carboniferous tetrapods the ossification of limbs, in conjunction with the limb proportion and thickness of cortical bone, are used for assessment of terrestriality (Long and

Gordon, 2004; Coates et al., 2008; Shubin et al., 2014). However, they do not always lead to the same conclusions about locomotor capability (Warren, 2007).

Observations about ossification and robusticity are qualitative descriptions in the early tetrapod literature and lack rigorous testing to associate them with known biological indicators of locomotor behaviour. Paleohistology and biomechanics are more biologically informed routes by which locomotor behaviour is assessed. A number of recent paleohistological investigations reveal patterns of internal anatomy in the limb bones of extant and extinct tetrapods, including thickness of cortical bone and long bone compactness, associated with various lifestyles (Laurin et al., 2004; Kriloff et al., 2008; Canoville and Laurin, 2009; Sanchez et al., 2010; Laurin et al.,

2011; Quemeneur et al., 2013; Sanchez et al., 2016; Molnar et al., 2017; Sanchez et al., 2017;

Molnar et al., 2018; Plasse et al., 2019). In extant tetrapods thicker layers of cortical bone are associated more with terrestrial animals, whereas a large proportion of trabecular bone is associated with an aquatic lifestyle (Kriloff et al., 2008; Canoville and Laurin, 2009; Laurin et al., 2011; Quemeneur et al., 2013; Plasse et al., 2019).

The biological response of trabecular bone to repetitive force, is to modify the bone structure, through modeling or remodelling, to maximize strength and minimize mass of the bone undergoing the repetitive force (Lieberman, 1997)This modelling and remodelling can be

8

quantified using the degree of anisotropy. Assessing forces from repetitive behaviour, such as ground reaction forces, on bone is used more often in paleoanthropological literature (Ryan and

Ketcham, 2002a, 2002b, 2005; Zeininger et al., 2016; Su and Carlson, 2017; Williams-Hatala et al., 2018). Similar techniques have not been used for studying the fin-to-limb transition in early tetrapods.

The extant phylogenetic bracket (EPB; Bryant and Russell, 1992; Witmer, 1995) is another method used to infer the presence of soft tissues and, by extension, potential locomotor capabilities. The EPB relies on extant bracket taxa to infer the presence of a soft tissue, with varying degrees of confidence, in a fossil taxon. The bracket taxa should be composed of two extant outgroups and an extant sister group to the fossil taxon of interest (Figure 1.1). The outgroup bracket for early tetrapods is composed of the lungfish, plus that group’s outgroup

(Figure 1.1) the coelacanths. The extant sister group to early tetrapods includes all living amphibians and amniotes, which are a diverse group of animals. Paleontologists use the EPB to examine the location and differentiation of musculature on the bone (Hutchinson and Gatesy,

2000; Hutchinson, 2001a, 2001b, 2002; Carrano and Hutchinson, 2002; Molnar et al., 2017,

2018). For early tetrapods the EPB requires comparisons between the fin of a lungfish (very highly derived among living animals) and coelacanth, and often a salamander or iguana. Using the EPB, locations of muscle attachment to fossil bone are inferred and patterns of muscle differentiation are observed across the fin-to-limb transition (Molnar et al. 2017; Diogo et al.

2016; Molnar et al. 2018). When applied to early tetrapods this method requires the assumptions that 1) phylogenies of early tetrapods are well resolved and, 2) osteological correlates are present and homologous from lungfish to amphibian and amniote taxa being used to bracket. This clearly

9

Figure 1.1. The outgroups and sister group to early tetrapods that are required to infer soft tissues using the extant phylogenetic bracket. involves multiple levels of assumptions and leads to tenuous conclusions when using the EPB in early tetrapods.

External morphological data from fossil material tells an incomplete story. Internal data that reflect biological responses of bone to forces exerted on them are useful to further support conclusions about locomotor capability. Observations of trends based on known biological and physiological responses of bone to behaviours associated with locomotor behaviour are critical for making new meaningful proxies. Additionally, such proxies need not rely on the EPB, or qualitative descriptions such as robusticity.

The purpose of this study is to use multiple qualitative and quantitative methods to analyse the tetrapod limb bones from Blue Beach. I will attempt to establish which morphotypes

10

were more aquatic or more terrestrial. I expect the diverse faunal assemblage at Blue Beach to reflect a combination of aquatic, amphibious, and terrestrial morphotypes. The animals at Blue

Beach similar to those suggested to be more amphibious, such as Pederpes (Clack and Finney,

2005), Proterogyrinus (Holmes, 1984) and Eoherpeton (Smithson, 1985), should have internal limb bone features that differentiate them from aquatic morphotypes such as those associated with Acanthostega (Coates, 1996) and Tulerpeton (Lebedev and Coates, 1995). Coming from near the base of Romer’s Gap, the are presumed to be close to the origin of terrestriality. The research impact of this study includes establishing the morphotype identity of new bone elements, and identifying variables signalling locomotor behaviour of animals at Blue Beach.

These variables can then be used to study fossils from elsewhere in order to better constrain statements regarding the evolution of terrestriality.

This study will 1) describe new tetrapod material from Blue Beach; 2) attempt to infer differentiation and direction of muscle insertions and associate them with a locomotor behaviour;

2) use extensive data from previous studies (Laurin et al., 2004, 2011; Kriloff et al., 2008;

Canoville and Laurin, 2009; Quemeneur et al., 2013) to infer lifestyle of Blue Beach animals from midshaft profiles; and 3) propose a new method to assess trabecular anisotropy over small regions of an entire bone and associate whole bone anisotropy frequency with locomotor behaviour.

Predictions

In the first chapter I will describe previously undescribed material from Blue Beach.

Recognition of new (or old) material can extend lineages both geographically and temporally.

However, the material first needs to be recognized and exposed to the paleontological community.

11

Muscle attachment scarring will be compared in the second chapter between extant amphibians, amniotes, and the Blue Beach material. I predict that if the mechanics of locomotion changed from operating in an aquatic environment to a terrestrial environment then; 1) muscle scarring patterns and orientations should shift from those more similar to extant aquatic taxa to more consistent with extant terrestrial taxa, 2) major bone features (such as trochanters) should shift from being present in locations consistent with extant aquatic animals to locations more consistent with extant terrestrial animals, and 3) Blue Beach limb bones should have muscle scarring comparable to a range of extant aquatic and extant terrestrial taxa.

The third chapter uses a computer program devised by Girindot and Laurin (2003) in which a two-dimensional slice at the midshaft of a long bone is used to create a quantitative profile. The variables from the program are plugged into a formula (see supplementary binary formula by Quemeneur et al. 2013) to assess aquatic, or not aquatic lifestyles of the animals associated with the bone. I predict that: 1) because a range morphotypes associated with amphibious Carboniferous (Whatcheerid, Proterogyrinus, Eoherpeton) and Devonian aquatic

(Acanthostega, Tulerpeton) tetrapods are preserved at Blue Beach the bones will reflect a range of terrestrial, amphibious, and aquatic lifestyles; 2) forms associated with the Carboniferous morphotypes will have terrestrial/amphibious profiles.

The fourth chapter will explore a new method using the bone analysis module of

Dragonfly (ORS Montréal). I will qualitatively assess where the strongest forces are acting on the Blue Beach fossil limb bones and create a vector map of the orientations of trabecular structure within each bone. Then I will use previously described variables such as bone volume to total volume (BV/TV) and anisotropy, in addition to variables unique to Dragonfly (ORS,

Montreal) such as cortical area fraction (CtA.TtA). I propose a new variable for assessing

12

locomotor behaviour of limb bones: frequency of anisotropy. I expect that more amphibious taxa such as Pederpes-like, Proterogyrinus-like and Eoherpeton-like tetrapods will have high trabecular orientation 1) in regions where musculature exerts force associated with terrestrial locomotion, and 2) across joints where gravitational and ground reactive forces are transferred along the limb. Finally, I expect aquatic taxa to show less or no concentrated regions of trabecular orientation, and if they do, I anticipate them to be along the midshaft and associated with musculature used to move the limbs in a lateral swimming motion.

The results and conclusions from chapters one through four will be summarized and their implications discussed in the fifth and final chapter.

Literature cited

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Barrell, J., E. Huntington, R. S. Lull, C. Schuchert, and L. L. Woodruff. 2011. The Evolution of

the Earth and Its Inhabitants; a Series of Lectures Delivered before the Yale Chapter of the

Sigma Xi during the Academic Year 1916-1917, by Joseph Barrell, Charles Schuchert,

Lorande Loss Woodruff, Richard Swann Lull, Ellsworth Huntingto, 3rd ed. Yale University

Press, New Haven, 1–235 pp.

Bennett, C. E., T. I. Kearsey, S. J. Davies, D. Millward, J. A. Clack, T. R. Smithson, and J. E. A.

Marshall. 2016. Early Mississippian sandy siltstones preserve rare vertebrate fossils in

seasonal flooding episodes. Sedimentology 63:1677–1700.

Boisvert, C. A., E. Mark-Kurik, and P. E. Ahlberg. 2008. The pectoral fin of Panderichthys and

the origin of digits. Nature 456:636–638.

13

Bryant, H. N., and A. P. Russell. 1992. The role of phylogenetic analysis in the inference of

unpreserved attributes of extinct taxa. Philosophical Transactions - Royal Society of

London, B 337:405–418.

Canoville, A., and M. Laurin. 2009. Microanatomical diversity of the humerus and lifestyle in

lissamphibians. Acta Zoologica 90:110–122.

Carrano, M. T., and J. R. Hutchinson. 2002. Pelvic and hindlimb musculature of Tyrannosaurus

rex (Dinosauria: Theropoda). Journal of Morphology 253:207–228.

Clack, J. A. 2001. Eucritta melanolimnetes from the Early Carboniferous of Scotland, a stem

tetrapod showing a mosaic of characteristics. Transactions of the Royal Society of

Edinburgh: Earth Sciences 92:75–95.

Clack, J. A. 2002. An early tetrapod from “Romer’s Gap”. Nature 418:72–6.

Clack, J. A. 2006. The emergence of early tetrapods. Palaeogeography, Palaeoclimatology,

Palaeoecology 232:167–189.

Clack, J. A., C. E. Bennett, D. K. Carpenter, S. J. Davies, N. C. Fraser, T. I. Kearsey, J. E. A.

Marshall, D. Millward, B. K. A. Otoo, E. J. Reeves, A. J. Ross, M. Ruta, K. Z. Smithson, T.

R. Smithson, and S. A. Walsh. 2017. Phylogenetic and environmental context of a

Tournaisian tetrapod fauna. Nature Ecology & Evolution 1.

Clack, J. A., R. L. Carroll, and H. Heatwole. 2000. Early Carboniferous tetrapods; pp. 1030–

1043 in Amphibian Biology. vol. 4 Palaeont.

Clack, J. A., and E. P. Ahlberg. 2004. A new stem tetrapod from the Early Carboniferous of

Northern Ireland. Recent Advances in the Origin and Early Radiation of Vertebrates 309–

14

320.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Clack, J. A., L. B. Porro, and C. E. Bennett. 2018. A Crassigyrinus-like jaw from the

Tournaisian (Early Mississippian) of Scotland. Earth and Environmental Science

Transactions of the Royal Society of Edinburgh 108:37–46.

Coates, M. I. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: postcranial anatomy,

basal tetrapod interrelationships and patterns of skeletal evolution. Transactions of the

Royal Society of Edinburgh: Earth Sciences 87:363–421.

Coates, M. I., and J. A. Clack. 1990. Polydactyly in the earliest known tetrapod limbs. Nature

347:66–69.

Coates, M. I., and J. A. Clack. 1991. Fish-like gills and breathing in the earliest known tetrapod.

Nature 352:234–236.

Coates, M. I., and J. A. Clack. 1995. Romer’s gap: tetrapod origins and terrestriality. Bulletin Du

Muséum National d’Histoire Naturelle 4:373–388.

Coates, M. I., M. Ruta, and M. Friedman. 2008. Ever Since Owen: Changing Perspectives on the

Early Evolution of Tetrapods. Annual Review of Ecology, Evolution, and Systematics

39:571–592.

Daeschler, E. B., N. H. Shubin, and F. A. Jenkins. 2006. A Devonian tetrapod-like fish and the

evolution of the tetrapod body plan. Nature 440:757–763.

Diogo, R., P. Johnston, J. L. Molnar, and B. Esteve-Altava. 2016. Characteristic tetrapod

15

musculoskeletal limb phenotype emerged more than 400 MYA in basal lobe-finned fishes.

Scientific Reports 6:37592.

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

Dunmore, C. J., T. L. Kivell, A. Bardo, and M. M. Skinner. 2019. Metacarpal trabecular bone

varies with distinct hand-positions used in hominid locomotion. Journal of Anatomy

235:45–66.

Georgiou, L., T. L. Kivell, D. H. Pahr, L. T. Buck, and M. M. Skinner. 2019. Trabecular

architecture of the great ape and human femoral head. Journal of Anatomy 234:679–693.

Germain, D., and M. Laurin. 2005. Microanatomy of the radius and lifestyle in amniotes

(Vertebrata, Tetrapoda). Zoologica Scripta 34:335–350.

Girondot, M., and M. Laurin. 2003. Bone Profiler: A tool to quantify, model, and statistically

compare bone-section compactness profiles. Journal of Vertebrate Paleontology 23:458–

461.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

Sciences 306:431–524.

Hutchinson, J. R. 2001a. The evolution of femoral osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:169–197.

Hutchinson, J. R. 2001b. The evolution of pelvic osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:123–168.

16

Hutchinson, J. R. 2002. The evolution of hindlimb tendons and muscles on the line to crown-

group birds. Comparative Biochemistry and Physiology - A Molecular and Integrative

Physiology 133:1051–1086.

Hutchinson, J. R., and S. M. Gatesy. 2000. Adductors, abductors, and the evolution of archosaur

locomotion. Paleobiology 26:734–751.

Inger, R. F. 1957. Ecological Aspects of the Origins of the Tetrapods. Evolution 11:373.

Kivell, T. L., R. Davenport, J. J. Hublin, J. F. Thackeray, and M. M. Skinner. 2018. Trabecular

architecture and joint loading of the proximal humerus in extant hominoids, Ateles, and

Australopithecus africanus. American Journal of Physical Anthropology 167:348–365.

Kriloff, A., D. Germain, A. Canoville, P. Vincent, M. Sache, and M. Laurin. 2008. Evolution of

bone microanatomy of the tetrapod tibia and its use in palaeobiological inference. Journal of

Evolutionary Biology 21:807–826.

Laurin, M., M. Girondot, and M.-M. Loth. 2004. The evolution of long bone microstructure and

lifestyle in lissamphibians. Paleobiology 30:589–613.

Laurin, M., A. Canoville, and D. Germain. 2011. Bone microanatomy and lifestyle: A

descriptive approach. Comptes Rendus - Palevol 10:381–402.

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Lieberman, D. E. 1997. Making Behavioral and Phylogenetic Inferences from Hominid Fossils:

Considering the Developmental Influence of Mechanical Forces. Annual Review of

Anthropology 26:185–210.

17

Lombard, E. F., and J. R. Bolt. 1995. A New Primitive Tetrapod, Whatcheeria Deltae, From The

Lower Carboniferous of Iowa. Palaeontology 38:471–494.

Long, J. A., and M. S. Gordon. 2004. The greatest step in vertebrate history: A paleobiological

review of the fish-tetrapod transition. Physiological and Biochemical Zoology 77:700–719.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, T. A., and M. R. Gibling. 1996. Stratigraphy and tectonic history of the Upper Devonian

to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology 32:13–38.

Milner, A. C., and W. Lindsay. 1998. Postcranial remains of Baphetes and their bearing on the

relationships of the Baphetidae (= Loxommatidae). Zoological Journal of the Linnean

Society 122:211–235.

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2017. Reconstructing pectoral

appendicular muscle anatomy in fossil fish and tetrapods over the fins-to-limbs transition.

Biological Reviews. 93:1077-1107

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2018. Evolution of Hindlimb Muscle

Anatomy Across the Tetrapod Water-to-Land Transition, Including Comparisons with

Forelimb Anatomy. Anatomical Record. doi: 10.1002/ar.23997.

Mossman, D. J., and R. G. Grantham. 2008. Eochelysipus horni, a new vertebrate trace fossil

from the Tournaisian Horton Bluff Formation, Nova Scotia. Atlantic Geology 44:69–77.

Panchen, A. L. 1985. On the Amphibian Crassigyrinus scoticus Watson from the Carboniferous

18

of Scotland. Philosophical Transactions of the Royal Society of London. Series B 309:505–

568.

Panchen, A. L., and T. R. Smithson. 1990. The pelvic girdle and hind limb of Crassigyrinus

scoticus (Lydekker) from the Scottish Carboniferous and the origin of the tetrapod pelvic

skeleton. Transactions of the Royal Society of Edinburgh: Earth Sciences 81:31–44.

Pierce, S. E., J. A. Clack, and J. R. Hutchinson. 2012. Three-dimensional limb joint mobility in

the early tetrapod Ichthyostega. Nature 486:523–526.

Plasse, M., E. Amson, J. Bardin, Q. Grimal, and D. Germain. 2019. Trabecular architecture in

the humeral metaphyses of non‐avian reptiles (Crocodylia, Squamata and Testudines):

Lifestyle, allometry and phylogeny. Journal of Morphology jmor.20996.

Quemeneur, S., V. de Buffrénil, and M. Laurin. 2013. Microanatomy of the amniote femur and

inference of lifestyle in limbed vertebrates. Biological Journal of the Linnean Society

109:644–655.

Romer, A. S. 1956. The Early Evolution of Land Vertebrates. Proceedings of the American

Philosophical Society 100:157–167.

Romer, B., and S. J. Godfrey. 1989. The Postcranial Skeletal Anatomy of the Carboniferous

Tetrapod Greererpeton burkemorani Romer, 1969. Philosophical Transactions of the Royal

Society B: Biological Sciences 323:75–133.

Ryan, T. M., and R. A. Ketcham. 2002a. The three-dimensional structure of trabecular bone in

the femoral head of strepsirrhine primates. Journal of Human Evolution 43:1–26.

Ryan, T. M., and R. A. Ketcham. 2002b. Femoral head trabecular bone structure in two

19

omomyid primates. Journal of Human Evolution 43:241–263.

Ryan, T. M., and R. A. Ketcham. 2005. Angular orientation of trabecular bone in the femoral

head and its relationship to hip joint loads in leaping primates. Journal of Morphology

265:249–263.

Sallan, L. C., M. I. Coates, J. Clack, L. C. Saltan, and M. I. Coatesa. 2010. End-Devonian

extinction and a bottleneck in the early evolution of modern jawed End-Devonian extinction

and a bottleneck in the early evolution of modern jawed vertebrates. Proceedings of the

National Academy of Sciences of the United States of America 107:10131–10135.

Sanchez, S., V. Dupret, P. Tafforeau, K. M. Trinajstic, B. Ryll, P. J. Gouttenoire, L. Wretman, L.

Zylberberg, F. Peyrin, and P. E. Ahlberg. 2013. 3D Microstructural Architecture of Muscle

Attachments in Extant and Fossil Vertebrates Revealed by Synchrotron Microtomography.

PLoS ONE 8:e56992.

Sanchez, S., P. Tafforeau, J. A. Clack, and P. E. Ahlberg. 2016. Life history of the stem tetrapod

Acanthostega revealed by synchrotron microtomography. Nature 537:408–411.

Sanchez, S., P. E. Ahlberg, K. M. Trinajstic, A. Mirone, and P. Tafforeau. 2012. Three-

dimensional synchrotron virtual paleohistology: A new insight into the world of fossil bone

microstructures. Microscopy and Microanalysis 18:1095–1105.

Sanchez, S., D. Germain, A. De Ricqlès, A. Abourachid, F. Goussard, and P. Tafforeau. 2010.

Limb-bone histology of temnospondyls: Implications for understanding the diversification

of palaeoecologies and patterns of locomotion of Permo-Triassic tetrapods. Journal of

Evolutionary Biology 23:2076–2090.

20

Shubin, N. H., E. B. Daeschler, and F. A. Jenkins. 2014. Pelvic girdle and fin of Tiktaalik roseae.

Proceedings of the National Academy of Sciences of the United States of America

111:893–899.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Smithson, T. R., and J. A. Clack. 2013. Tetrapod appendicular skeletal elements from the Early

Carboniferous of Scotland. Comptes Rendus - Palevol 12:405–417.

Smithson, T. R., and J. A. Clack. 2018. A new tetrapod from Romer’s Gap reveals an early

adaptation for walking. Earth and Environmental Science Transactions of the Royal Society

of Edinburgh 108:89–97.

Smithson, T. R., S. P. Wood, J. E. A. Marshall, and J. A. Clack. 2012. Earliest Carboniferous

tetrapod and arthropod faunas from Scotland populate Romer’s Gap. Proceedings of the

National Academy of Sciences of the United States of America 109:4532–7.

Su, A., and K. J. Carlson. 2017. Comparative analysis of trabecular bone structure and

orientation in South African hominin tali. Journal of Human Evolution 106:1–18.

Tibert, N. E., and D. B. Scott. 1999. Ostracodes and Agglutinated Foraminifera as Indicators of

Paleoenvironmental Change in an Early Carboniferous Brackish Bay, Atlantic Canada.

PALAIOS 14:246.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

21

Ward, P., C. Labandeira, M. Laurin, and R. A. Berner. 2006. Confirmation of Romer’s Gap as a

low oxygen interval constraining the timing of initial arthropod and vertebrate

terrestrialization. Proceedings of the National Academy of Sciences 103:16818–16822.

Warren, A. 2007. New data on Ossinodus pueri, a stem tetrapod from the Early Carboniferous of

Australia. Journal of Vertebrate Paleontology 27:850–862.

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Waters, C. N., R. A. Waters, W. J. Barclay, and J. R. Davies. 2009. A Lithostratigraphical

Framework for the Carboniferous Successions of Northern Great Britain (Onshore). 174 pp.

Williams-Hatala, E. M., K. G. Hatala, M. Gordon, A. Key, M. Kasper, and T. L. Kivell. 2018.

The manual pressures of stone tool behaviors and their implications for the evolution of the

human hand. Journal of Human Evolution 119:14–26.

Witmer, L. M. 1995. The Extant Phylogenetic Bracket and the importance of reconstructing soft

tissues in fossils; pp. 19–33 in Jeff Thomason (ed.), Functional morphology in vertebrate

paleontology. Cambridge University Press.

Zeininger, A., B. A. Patel, B. Zipfel, and K. J. Carlson. 2016. Trabecular architecture in the StW

352 fossil hominin calcaneus. Journal of Human Evolution 97:145–158.

22

Chapter 2: Descriptions of new tetrapod limb elements from Blue Beach, Nova Scotia Introduction

In 1841 Sir William Logan first observed trackways in the Horton Bluff formation. Sir

William Dawson wrote about trackways from the site in the later 1880s, but bones of tetrapods were not recovered until Baird and Carroll worked the site in the 1960s (Carroll et al., 1972;

Sues et al., 2013). Most previous discussions of Blue Beach tetrapods simply note that isolated material is preserved (Clack et al., 2000; Hall, 2007). Only recently has such material been described from other early Carboniferous sites (Lombard and Bolt, 1995; Clack, 2001, 2002;

Warren and Turner, 2004; Clack and Finney, 2005; Warren, 2007). The tetrapod material from

Blue Beach consists of disarticulated, isolated elements, mostly from the limbs. One recent effort to describe some of the isolated limb bones approached it from the perspective of morphotypes

(Anderson et al., 2015), because enough articulated skeletal material from the Earliest

Carboniferous (Tournaisian and Visean) has been described to aid in assigning the isolated elements to higher taxonomic identifications of such reference material (i.e., Pederpes, Clack and Finney, 2005; Whatcheeria, Lombard and Bolt, 1995; Ossinodus, Warren and Turner, 2004;

Eucritta, Clack, 2001). Efforts to increase sampling of the Tournaisian are also currently underway in Scotland (Clack et al., 2018; Smithson and Clack, 2018; Otoo et al., 2019).

The material that follows is of undescribed and unfigured tetrapod fossil remains from the base of Romer’s Gap at Blue Beach Nova Scotia.

Abbreviations

Blue Beach Fossil Museum, Hantsport, Nova Scotia, Canada (BBFM); Nova Scotia

Museum of Natural History, Halifax, Canada (NSM); Redpath Museum, McGill University,

23

Montreal, Canada (RM); Yale Peabody Museum of Natural History, New Haven, Connecticut,

USA, (YPM).

ab, adductor blade; ac, adductor crest; cn, cnemial crest; cp, central prominence; der, dorsal elongate crest; ff, fibular facet; fo, fibular fossa; ft, fourth trochanter; icf, intercondylar fossa; it, internal trochanter; itf, intertrochanteric fossa; mr, mesial ridge; pa, popliteal area; tf, tibial facet.

Geological context

Blue Beach represents one of the earliest known tetrapod bearing formations of the

Tournaisian. Based on palynozonation it dates late Tn2 (late middle-Tournaisian) to early Tn3

(early late-Tournaisian) subzones, (Utting et al., 1989). This is contemporaneous to slightly older than the tetrapod bearing CM (claviger–macra) zone in Europe, which dates to the early tn3 subzone (Utting et al., 1989). Scottish sites of this age have recently revealed articulated tetrapod material (Smithson and Clack, 2013, 2018; Kearsey et al., 2016; Clack et al., 2018).

However, current work re-examining the Tournaisian localities in Scotland suggests sites from the Ballagran formation may extend below the CM to the PC (pretiosus-clavata) palynozone, so articulated tetrapod material from these formations may be older (Waters et al., 2009; Clack et al., 2017, 2019; Otoo et al., 2019).

The Blue Beach depositional environment was first interpreted as a non marine high energy environment (Martel and Gibling 1991). Later, a periodic marine influence was suggested

(Martel et al., 1993; Martel and Gibling, 1996). The Blue Beach depositional environment is now known as a marginally marine location with at least one fully marine transgression (Tibert and

Scott, 1999; Mansky and Lucas, 2013). The tetrapod material is disarticulated and has signs of

24

exterior weathering. This suggests the animals were transported a short distance before deposition and fossilization (Mansky and Lucas, 2013).

Blue Beach is located between Hantsport and Avonport, Nova Scotia, Canada, where record breaking high tides of over 10m constantly erode and expose new material. Much of the material is preserved exceptionally well in three dimensions despite high wave and tidal action both at the time of deposition and collection. This is critical for analyses of external and internal structures. External pits, scars, ridges, and foramina are visible on the bone surface and can aid in identification of the animal, it’s physiology, and lifestyle.

Material and Methods

Many specimens were prepared by Sonja Wood of the BBFM prior to being borrowed.

Any unprepared or partially prepared specimens were fully prepared out of the matrix by KL using Paleotools microjacks 1-3, a Chicago Pneumatics air scribe, and pin vice. The elements were reinforced during preparation with polyvinyl acetate and carbowax. NSM.007.GF.004.733 was prepared by Alan Lindoe (University of Alberta). To assess if flattening was a taphonomic or true biological feature of the element the limb bones were scanned in the Xradia Versa 510 at the University of Calgary with settings of 140kV, 72µA, no filter, and magnification 0.4x.

Results

Humeri

Anderson et al. (2015) previously described the humeri YPM.PU.23545 and

YPM.PU.20754, which were borrowed from the Yale Peabody Museum (YPM; New Haven,

Connecticut).

25

Ulna

NSM.007.GF.004.733

The element (Figure 2.1) is a right ulna and measures 34.3mm in length. It is similar in size to the ulna of Pederpes (Clack and Finney, 2005) from the tip of the prominent olecranon process to the distal surface. This element is heavily worn, but the internal structure reveals minimal crushing. The proximal region of the bone is dominated by the olecranon process, which extends approximately 8mm beyond the rest of the proximal articulating surface. The lateral and mesial edges are concave. A ridge (mr; Figure 2.1 C) runs along the anterior edge of the flexor surface of the element. No other prominent rugosity or features are distinguishable on the bone surface. The element is teardrop shaped in proximal view (Figure 2.1 E) and slightly U-shaped in distal view (Figure 2.1 F). The element is more gracile than ulnae of Acanthostega and

Ichthyostega, but not as elongate as those of Archeria or Greerepeton (Romer, 1957; Romer and

Godfrey, 1989; Coates, 1996; Jarvik, 1996). The anterior edge of the ulna is more concave than in Tulerpeton (Lebedev and Coates, 1995), and the waist at the midshaft is more distinct. The element may be comparable to the ulna of Crassigyrinus (Panchen, 1985), but it has a narrower midshaft with respect to the proximal and distal ends than the ulna figured in Panchen.

Additionally, the ulna of Crassigyrinus does not have a prominent olecranon process preserved.

The ulna of Pederpes also lacks a prominent olecranon process and is less curved along the anterior edge than the ulna of Tulerpeton or NSM.007.GF.004.733 (Lebedev and Coates, 1995;

Clack and Finney, 2005).

26

Figure 2.1. Micro computed tomography (µCT) images of NSM.007.GF.004.733: right ulna. Tulerpeton- like ulna in (A) extensor, (B) anterior, (C) flexor, (D) posterior, (E) proximal and (F) distal view. mr, mesial ridge. Scalebar 10mm. The element most closely resembles the ulna of Tulerpeton for the reasons outlined above.

Radii

NSM.014.GF.036.002

The element (Figure 2.2 A-F) is 32.7mm from proximal to distal end. It is elongate and the flexor, extensor, anterior, and posterior edges are minimally convex approaching straight.

The midshaft is nearly square in outline (seen in transaxial microcomputed tomography

(microCT) image slices), becoming cylindrical towards the proximal end. The flexor surface displays the most rugosity (Figure 2.2 C). It has striations oriented proximodistally radiating from the long furrow located near the distal midshaft. Both flexor and extensor surface are slightly concave. The flexor surface is bound by a sharp anterior ventral radial crest (Figure 2.2

C, D) and a sharply angled posterior edge. The edge between the posterior and extensor surfaces has a distal prominence with a convex curve in extensor view. The proximal articulating surface

27

Figure 2.2. µCT images of NSM.014.GF.036.02, right Acanthostega-like radius (A-F) and NSM.007.GF.004.767, left radius (G-L) in (A, G) extensor, (B, H) posterior, (C, I) flexor, (E, K) proximal and (F, L) distal view. vrc, ventral radial crest. Scalebar 10mm. is a hollow, circular, concave shape and the distal articulating surface is flat and D-shaped.

MicroCT scans reveal the element is excellently preserved in three-dimensions.

Based on the circular proximal articulating surface, thin ventral radial crest, and acutely angled edges along a straight shaft, the element most closely resembles the radius of

Acanthostega.

NSM.007.GF.004.767

This element, like NSM.014.GF.036.002 (Figure 2.2 G-L), has a nearly square profile at the midshaft seen in transaxial microCT slices. It is slightly smaller at 29.0mm. The ventral

28

radial crest is a thin flat projection where the flexor and anterior surfaces meet (Figure 2.2 G, I,

J). No rugosity is discernable on the ventral surface (Figure 2.2 I). The flexor surface is slightly concave, compared to the convex extensor surface. The proximal articulating surface is more cylindrical than the distal end in (transaxial sections); however, the distal end in flexor view is more rounded. The proximal articulating surface is sub oval (Figure 2.2 K), and the distal articulating surface crescent-shaped (Figure 2.2 L).

The element also appears to most closely resemble the radii of Acanthostega, or possibly the more gracile Archeria (Romer, 1957). Internal analysis shows the element is well preserved in three dimensions. Interestingly, it has a very thick cortical layer of bone, whereas

NSM.007.GF.004.767 does not.

Femora

NSM.005.GF.045.048-A

This is a left femur (Figure 2.3 A-F) that is well preserved in three dimensions. It is

60.6mm long and exhibits heavy wear on the adductor blade. No trochanters are clearly visible as a result of the wear. What remains of the adductor blade is separated from the midshaft by a vertical crack. The specimen appears crushed at the distal end, but the midshaft has excellent three-dimensional preservation revealed in transaxial microCT slices. The intertrochanteric fossa is located proximo-ventrally and is a teardrop shape that is highly rugose. Rugose pits are smaller, and closer to circular at the proximal curved end of the teardrop shape. They are deeper and more elongate at the V-shaped end of the intertrochanteric fossa near the midshaft. Distally on the ventral surface there is a sub-triangular region of rugosity within the popliteal area. The popliteal area has more elongate scarring than that observed within the intertrochanteric fossa.

The dorsal surface is smooth and cylindrical at the midshaft. No prominent rugosity or processes

29

Figure 2.3. µCT images of NSM.005.GF.045.048-A, left embolomerous femur (A-F), and RM.20.5222, a right Acanthostega-like) femur (G-L) in (A, G) extensor, (B, H) anterior, (C, I) flexor, (D, J) posterior, (E, K) proximal, and (F, L) distal views. ab, adductor blade; ac, adductor crest; icf, intercondylar fossa; itf, intertrochanteric fossa; pa, popliteal area. Scalebar 10mm. are present. The anterior surface has numerous small pits converging at the midshaft. The posterior surface is highly concave compared to the anterior. The proximal articulating surface appears oval, is unfinished (lacks a continuous ossified layer of bone) and reveals the internal trabecular structure. The distal articulating surface is an elongate oval with a distinct ridge on the ventral surface of the midline. The ridge has a corresponding concavity (intercondylar fossa) on the dorsal side of the distal surface. The ridge does not clearly divide the articulating surface. In dorsal view there is no discernable bulbous swelling as in the femur of Tulerpeton.

The element is similar to the description of Femur Type 2 from Anderson et al. (2015), which would suggest an embolomerous animal. The posterior profile in ventral view is extremely

30

convex with respect to the anterior profile in ventral view. Additionally, the dorsal profile appears flat like Femur Type 2.

RM.20.6711

This femur was borrowed from the Redpath Museum (Montréal) for this study. The ventral surface is figured in Clack and Carroll (2000); additional surfaces are illustrated in Figure

2.3 G-H (this work).

A teardrop-shaped intertrochanteric fossa covers most of the proximal end of the ventral surface. The intertrochanteric fossa has smaller pits near the proximal convex end and large elongate pits near the V-shaped end at the midshaft. The adductor blade is worn, although not to the same degree as in NSM.005.GF.045.048-A. The adductor crest provides a posterior edge to the popliteal area which also contains small circular and deep elongate pits. The dorsal surface is smooth and cylindrical. The dorsal profile in anterior or posterior view is sigmoidal, similar to the femur of Acanthostega (Coates 1996). Unlike in NSM.005.GF.045.048-A this element does not appear flattened at the proximal or distal ends. The anterior surface is less concave compared to the posterior surface, but not to the same extent as in NSM.005.GF.045.048-A. The adductor blade of RM.20.6711 does not extend distally as far as in the femur of Acanthostega. Like

NSM.005.GF.045.048-A there are small pits located all over the anterior surface. The posterior surface is smooth. The proximal articulating surface is oval, and the distal articulating surface

V-shaped.

The element shares many similarities to the femur of Acanthostega but is quite different to femur morphotype 4 described in Anderson et al. (2015). The general shape and size are similar to Blue Beach femur NSM.005.GF.045.048-A, but details such as the sigmoidal dorsal

31

Figure 2.4. µCT images of NSM.007.GF.004.630 a right Tulerpeton-like femur in (A) extensor, (B) anterior, (C) flexor, (D) posterior, (E) proximal, and (F) distal view. ab, adductor blade; ac, adductor crest; cp, central prominence; ff, fibular facet; fo, fibular fossa; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm. surface, and less convex posterior surface suggest and association of RM.20.6711 with

Acanthostega.

NSM.007.GF.004.630

A left femur 65.5mm long (Figure 2.4), is excellently preserved in three dimensions with negligible crushing. The element has a narrower waist than the previously described femora from this study. The proximal end is bulbous in shape and there is approximately 40 degrees of rotation between the proximal and distal articulating surfaces. A crack bisects the element at the midshaft, but what remains of the adductor blade is attached unlike in NSM.007.GF.004.630.

The dorsal surface appears rugose although it is difficult to tell if it is due to preparation, wear, or is a real biological feature. The proximal half of the dorsal surface is very round and moderately wide. There is a ridge similar to the central prominence (cp) described in the femur of Tulerpeton (Lebedev and Coates, 1995). The element is cylindrical at the midshaft before two distal condyles emerge. The condyles extend an equal distance dorsally. The articulating surface of the anterior condyle projects slightly anteriorly (i.e., not in a horizontal plane, approximately

32

45 degrees from horizontal) whereas the posterior condyle projects distally (i.e., horizontal). The intercondylar fossa is shallow and broad. It extends less than a third of the way up the bone and ends near the distal articulating surface. The ventral surface is also rugose, but much of it looks to be from wear, as exposed trabecular bone is visible. Most of the cortical bone of the distal and posterior edges of the intertrochanteric fossa is worn, exposing trabecular bone beneath. The anteroproximal boundary has a small border of ossified cortical bone (finished bone; arrow in

Figure 2.4 B). The proximal portion of the intertrochanteric fossa is composed of relatively smooth finished bone. The distal teardrop-shaped end is worn exposing the trabecular bone beneath. Most of the adductor blade is absent. The portion of the blade that remains is a triangular projection extending ventrodistally. Exposed trabecular bone suggests the proximal part of the adductor blade broke off rather than the triangular projection reflecting a biological feature. No trochanters are distinguishable. The adductor crest is composed of finished bone. It runs from the distal edge of the adductor blade towards the distal articulating surface and posterior of the midline. The adductor crest and adductor blade meet at right angles with the blade projecting ventrally. The popliteal area is rugose and posteriorly lined by the adductor crest. Large deep elongate foramina on are located where the popliteal area meets the adductor crest. The finished bone preserved on the anterior surface of the midshaft has numerous small pits (see Figure 3.5). The anterior surface is less concave than the posterior surface. The posterior surface preserves little finished bone with the exception of the midshaft, which preserves a small area covered with finished bone. Striations radiate proximally from the posterior surface of the midshaft, but they are obscured by wear. The proximal articulating surface viewed proximally is teardrop shaped. The surface is composed of exposed trabecular bone. The distal articulating surface when viewed dorsal side facing up is a deltoid shape. The intercondylar fossa extends

33

Figure 2.5. µCT images of YPM.PU.20103 a right Tulerpeton-like/whatcheeriid femur (A-F) and NSM.005.GF.045.042 a left Tulerpeton-like femur (G-L) in (A,G) extensor, (B,H) anterior, (C,I) flexor, (D,J) posterior, (E,K) proximal, and (F,L) distal views. scalebar 10mm. distally from the dorsal surface providing a small amount of finished bone between the tibial and fibular facets. However, ventrally the facets are connected by exposed trabecular bone.

The narrow cylindrical waist and bulbous proximal end make it unlike any other femur I have observed. It shares a proximodorsal tuberosity with the femur of Tulerpeton, a remarkably thin midshaft with the femur of Ossinodus, and is most like Blue Beach specimen

YPM.PU.23545 (Type 1 femur).

34

NSM.005.GF.045.042

This is a 68.25mm long left femur (Figure 2.5 G-L, and Figure 2.6 E-H) partially prepared by SW at the BBFM before completed by KL at the University of Calgary. In dorsal view the element looks well-preserved; however, preparation of the ventral surface revealed significant crushing of the specimen. Much of the anatomy, including a well-preserved adductor blade remains. However, the ventral surface is crushed into the marrow cavity and the cortical bone is cracked. Overall the element is minimally waisted with very little expansion or broadening at the proximal and distal ends. Axial twist is small but discernable, similar to the femur of Tulerpeton (Lebedev and Coates, 1995). A slight elongate but small ridge is present along the posterior edge of the proximal half of the dorsal surface (der; dorsal elongate ridge).

The proximal two thirds of the dorsal surface appear free of cracking and preserve irregularly shaped pits across the surface. The distal third exhibits much cracking and crushing but remains unbroken. The intercondylar fossa is prominent, if made deeper by crushing. On the dorsal surface the triangular tip of the fossa extends to the posterior end of the adductor blade on the ventral surface. The fibular facet extends further distally than the tibial. Additionally, the tibial facet projects ventroanteriorly approximately 45 degrees offset from the horizontal, unlike the fibular facet which angles ventroposteriorly. The ventral surface has been crushed but the adductor crest, blade and the ridge creating the posterior edge of the intertrochanteric fossa are composed of finished bone and remain attached to the shaft. The adductor blade is robust, consisting of finished bone. It projects ventrally and remains a subrectangular from proximal to distal ends. The adductor blade runs approximately one third the length of the bone. Remains of the intertrochanteric fossa are distinct and are lined anteriorly and posteriorly with finished bone.

35

Figure 2.6. µCT images (meshes) of YPM.PU.20103 (A-D) and NSM.005.GF.045.042 (E-H) in (A,E) extensor, (B,F) anterior, (C,G) flexor and, (D,H) posterior views. ab, adductor blade; ac, adductor crest; der, dorsal elongate ridge; ff, fibular facet; fo, fibular fossa; icf, intercondylar fossa; it, internal trochanter; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm. The curved proximal boundary of the tear drop shaped fossa is unfinished. Trabecular structure is visible from the tip of the fossa to the proximal tip of the element. The popliteal area retains rugose large pits. These pits extend into the bone, rather than remain on the bone surface. The

36

anterior surface is moderately well preserved, but no rugosity or pitting is observed. The posterior surface is cracked, and no rugosity is observed. Both anterior and posterior surfaces are slightly concave. The proximal articulating surface appears to be composed of unfinished bone.

The unfinished bone extends at a shallow angle ventrally towards the proximal end of the intertrochanteric fossa. It is unclear if this is the result of crushing, wear, or a true feature of the bone. The proximal articulating surface is a sub oval D-shape. The distal articulating surface is also composed of exposed trabecular bone that extends at a shallow angle towards the ventral surface. The extension of the distal articulating surface to the ventral surface of the bone is considerably less than that described for the proximal end. The distal articulating surface forms a deltoid shape common to many early tetrapod femora. The anterior edge of the deltoid shape extends further than the posterior edge, which has a comma-like shape. The tibial and fibular facets are not separated by finished bone.

The squared off adductor blade dorsal ridge (perhaps related to the cp), and minimal waisting and flaring make this femur most closely resemble a Type 1 Tulerpeton-like femur.

NSM.005.GF.045.044

This specimen preserves the distal two thirds of a femur (Figure 2.7 A-E) measuring

50.05mm in length. The element appears to be cut diagonally at the distal end of the midshaft; it is cut at such an angle so that most of the distal adductor blade remains. The midshaft is more waisted and the distal end of the bone appears broader than in the femora of both Tulerpeton

(Lebedev and Coates, 1995) and Acanthostega (Coates, 1996). The adductor blade has a bifurcation at the proximal end and several foramina are located within. This bifurcation appears consistent with the distal edge of the intertrochanteric fossa.

37

What remains of the dorsal surface is moderately eroded. The distal condyles are slightly better preserved. The anterior condyle projects anteriorly whereas the posterior condyle projects distally. The anterior condyle has a sigmoidal curve in dorsal view, a feature not observed in any of the femora described above. No similar curve is noted in the femora of Tulerpeton,

Whatcheeria, or Ossinodus. Compared to NSM.007.GF.004.630 and NSM.005.GF.045.048-A the posterior condyle of NSM.005.GF.045.044 projects further distally in relation to the anterior condyle. It also has a smaller sigmoidal curve in dorsal view. The anterior condyle is much larger and broader than the posterior condyle. The intercondylar fossa projects less proximally compared to the other femora; however, erosion of the cortical layer obscures the proximal tip of the fossa. The intercondylar fossa in NSM.005.GF.045.044 extends ventrally at the distal end of the dorsal surface, which creates a trough when the element is viewed distally. The trough extends approximately halfway across the distal articulating surface. The most proximoventral region of the adductor blade appears to have remains of the intertrochanteric fossa. The distal end of the blade (4th trochanter) appears unfinished; however, the unfinished region is nearly perfectly oval similar to the 4th/internal trochanter of the femur of Pederpes (Clack and Finney,

2005), or the 4th trochanter of the femur of Ossinodus. The meeting of the adductor blade and adductor crest forms a gentle concave curve (seen best in Figure 2.8 D). This is unlike the L- shaped curve where the adductor blade meets the adductor crest described in most of the above femora. The popliteal area is a triangular depression with many foramina (large irregularly shaped pits) perforating the distal portion. Little rugosity appears on the anterior or posterior surfaces. Both surfaces have concave profiles. The fibular fossa is more depressed than in any of the Blue Beach femur foramina considered for this analysis and is a sub rectangular shape with

38

Figure 2.7. µCT images of NSM.005.GF.045.044, distal end of a left Ossinodus-like femur (A-E), and NSM.005.GF.045.047 proximal (K-O) and distal (F-J) segments in (A,F,K) extensor, (B,G,L) anterior, (C,H,M) flexor, (D,I,N) posterior, (E,J) distal and (O) proximal views. ab, adductor blade; ac, adductor crest; cp, central prominence; ff, fibular facet; fo, fibular fossa; ft, fourth trochanter; itf, intertrochanteric fossa; pa, popliteal area; tf, tibial facet. Scalebar 10mm. perforating foramina throughout the depression. The distal articulating surface is deltoid-shaped, a feature shared by many of the Blue Beach femora, and most early tetrapod femora.

The narrow waist, projecting oval shape of the fourth trochanter, and sharp angle between the fibular and tibial facets indicate that this was from an Ossinodus-like animal.

NSM.005.GF.045.047

This element is broken into two pieces (Figure 2.7 F-O), and a 5-10mm section between the two is missing. The proximal piece is 25.4mm long and the distal piece measures 38.3mm.

39

The adductor blade is heavily eroded, as is the distal end of the ventral surface. Each “half” is well preserved in three dimensions.

The proximal segment (Figure 2.7 K-O) is composed of a small amount of the proximal tip of the adductor blade, intertrochanteric fossa and proximal articulating surface. The dorsal surface shows signs of heavy wear and exposed trabecular bone. The dorsal surface is rounded with the apex of the curve located posterior to the midline. On the ventral surface the intertrochanteric fossa is surrounded by finished bone on the anterior, distal and posterior edges.

There are small pits and foramina located within. The proximal articulating surface is oval shaped.

The dorsal surface of the distal end (Figure 2.7 F-J) is waisted at the midshaft and distally flares out more than in the femur of Tulerpeton (Lebedev and Coates, 1995). The anterior condyle flares anteriorly 45 degrees from horizontal while the posterior condyle extends distally and faces more posteriorly. This is like the configuration observed in NSM.005.GF.045.044. The intercondylar fossa does not extend on the dorsal surface to a location comparable with the adductor blade on the ventral surface. The fossa is obscured by damage at the posterior edge of the distal end. The most concave portion of the intercondylar fossa extends to the distal articulating surface, partially separating the tibial facet from the fibular facet. The distal end preserves most of the adductor blade and crest. The blade is heavily worn so that is consists of almost entirely unfinished bone. A sharp ridge where the blade meets the adductor crest is composed of cortical bone with no exposed trabecular bone. Distally the crest and popliteal areas are eroded. Striations radiate from the midshaft of the bone to the distal end of the adductor crest, the fibular facet, and anterodistal edge of the bone. The distal articulation surface is deltoid to V-

40

shaped. This element appears similar to the Ossinodus-like femur NSM.005.GF.045.044 because of the thin waist, and angle between fibular and tibial facets.

Previously described

The femur YPM.PU.23550 is described in Anderson et al. (2015), and femora

YPM.PU.20103 (Figure 2.5 A-F, 6 A-D, this study) and RM.20.6711 (Figure 2.3 G-K, this study) are figured but not described in Clack and Carrol (2000). All three femora were borrowed from their respective institutions for this study. Recent investigations in Scotland have revealed elements very similar to the Blue Beach fauna. YPM.20.20103 is remarkably similar in size and shape to the femur NHMUKR 5601 from Inchkeith (Smithson and Clack, 2013), although the

Inchkeith femur appears to be lacking an adductor blade.

A note on femur YPM.PU.20103

Warren and Turner (2004) observed that the femur of Ossinodus most closely resembles

YPM.PU.20103 and that of Tulerpeton. Anderson et al. (2015) suggested that YPM.PU.23550 is a less crushed morphotype of YPM.PU.20103. However, microCT scans reveal the trabecular structures of both elements are crushed, but the cortical layers are well-preserved. I suggest that both share similarities with the femora of Tulerpeton and Ossinodus/Whatcheeria and that

YPM.PU.23550 remain as a Type 1 tulerpetontid morphotype, while YPM.PU.20103 be assigned a new, more whatcheeriid-like morphotype (Femur Type 5).

Fibulae

NSM.005.GF.045.061

This element clearly resembles the fibula of Pederpes (Clack and Finney, 2005).

NSM.005.GF.045.061 (Figure 2.8 A-F) is 37.6mm long measured between proximal and distal

41

articulation surfaces. The element is waisted, and the distal end more anterodistally expanded than the proximal end. The fibula appears flattened along the dorsoventral plane; however, microCT scans reveal it is well preserved in three dimensions. The entre posterior edge of the flexor surface has a ridge curving concavely (pr; Figure 2.8 C). The flexor surface has minimal rugosity radiating posterodistally at its distal end. At the proximoanterior end of the ridge there are fewer but wider pits resembling foramina. The extensor surface is heavily worn but retains elongate oval foramina radiating from the mid-line of the bone towards the anterior and distal edges. Like the fibula Pederpes (Clack and Finney, 2005), this element has no torsion. The proximal articulating surface dips down into the flexor surface, while the distal articulating surface has an upward lip near the anterior edge of the flexor surface. The proximal articulating surface is teardrop shaped while the distal articulating surface is a flattened elongate oval. The element is extremely similar to the Pederpes fibula figured and described Clack and Finney

(2005).

NSM.005.GF.045.063

This fibula (Figure 2.8 G-L) is shorter than NSM.005.GF.045.061 (30mm), more flattened, and lacks a posterior ridge on the flexor surface. The element is waisted, and the distal end is more flared than the proximal end. The proximal articulating surface dips distally extending onto the flexor surface as in NSM.005.GF.045.061 and the fibula of Pederpes (Clack

42

Figure 2.8. µCT images of NSM.005.GF.045.061 left Pederpes-like fibula (A-F) in and left fibula NSM.005.GF.045.063 (G-L) in (A,G) extensor, (B,H) posterior, (C,I) flexor, (D,J) anterior, (E,K) proximal, and (F,L) distal views. Scalebar 10mm. and Finney, 2005). The flexor surface has very large pits that look like several large foramina radiating anteriorly from the midshaft. The extensor surface is most rugose at the anterior and posterior edges. The element has a slight triangular shaped depression in the distoanterior edge of the extensor surface. The anterior edge is more concave than NSM.005.GF.045.061, but this may be the result of flattening. The proximal articulating surface has a tri-lobed shape. The distal articulating surface is elongate, sub rectangular, and has concave flexor and extensor edges. The outline of the element resembles the fibula shown but not described in Lombard and Bolt (1995; plate 1). It is shorter and wider than the fibulae of both Pederpes and Tulerpeton.

43

NSM.005.GF.045.112

This Crassigyrinus-like fibula is described in Lennie, Mansky, and Anderson (In Press).

It appears most similar to the fibula of the Carboniferous and potentially secondarily aquatic tetrapod Crassigyrinus scoticus (Panchen, 1985; Panchen and Smithson, 1990; Carroll, 1992a;

Herbst and Hutchinson, 2019). NSM.005.GF.045.112 has less axial twist than the fibula of

Crassigyrinus and may represent a morphotype less specialized for aquatic locomotion.

Tibiae

RM.20.5222

This tibia (Figure 2.9 A-F) appears Proterogyrinus/embolomere-like. It is 33.55mm long and well preserved in three-dimensions. The flexor surface of the element is rugose near the centre and the rugosity lessens at the peripheries in the proximal and distal directions. Unlike

NSM.005.GF.045.320 the flexor surface of this element has a Y-shaped crest observed in the tibiae of Tulerpeton, Ossinodus, Proterogyrinus and NSM.014.GF.36.004 (Y; described as a proximoposterior to distoanterior ridge along the flexor surface; (Holmes, 1984; Lebedev and

Coates, 1995; Warren and Turner, 2004; Anderson et al., 2015). The posterior edge is concavely shaped like NSM.014.GF.036.003, NSM.014.GF.036.004 (Anderson et al., 2015), and the tibiae of Ossinodus (Warren and Turner, 2004), Eoherpeton (Smithson, 1985), and Proterogyrinus

(Holmes, 1984) rather than notched as in NSM.005.GF.045.320. The anterior surface forms sharp edges where it meets both flexor and extensor surfaces. It becomes rounded at the distal end, and the whole surface is sigmoidal in anterior view. The flat anterior surface bears a ridge comparable to the cnemial process of NSM.005.GF.045.320 and the tibia of Ossinodus (Warren and Turner, 2004), or the rugose ridge described by Holmes (1984) in the tibia of

Proterogyrinus. In proximal view the element is sub triangular to elongate oval. The distal

44

Figure 2.9. µCT images of left tibiae RM.20.2555 (A-F) and NSM.005.GF.045.320 (G-L) in (A,G) extensor, (B,H) anterior, (C,I) flexor, (D,J) posterior, (E,K) proximal and (F,L) distal views. Scalebar 10mm articulating surface approaches the L-shape described in Tulerpeton, Acanthostega, and

Ossinodus (Lebedev and Coates, 1995; Coates, 1996; Warren and Turner, 2004).

The element is perhaps most similar to the tibiae of Eoherpeton or Proterogyrinus but it lacks torsion and is less gracile.

NSM.005.GF.045.320

This tibia (Figure 2.9 G-L) measures 32.5mm from proximal to distal articulation surface.

This element is flattened in the flexor/extensor plane. In flexor view the element has rugose ridges extending in a radial fan-like pattern from the middle of the anterior edge. The extensor

45

surface has a prominent cnemial crest on the anterior edge near the mid-line of the element, similar to the tibiae of Ossinodus (Warren and Turner, 2004) and Proterogyrinus (Holmes,

1984). There are small pits covering the entire extensor surface of the bone with elongated depressions radiating out proximally and distally. The pits are small and deep near the midshaft, becoming shallower and steeply angled near the proximal and distal ends. Viewed anteriorly the anterior surface has a sigmoidal curve between proximal and distal ends (Figure 9 H). There is no rotation between the proximal and distal articulating surfaces. The posterior profile is convex except for a concave nearly 90 degree notch (M-shaped?) at the midshaft (seen in Figure 2.9 G,

I). The notch is much less pronounced than the waisting observed in the tibiae of Pederpes

(Clack and Finney, 2005) or Ossinodus (Warren and Turner, 2004). The proximal articulating surface is an irregular oval shape, as is the distal articulating surface. The flattening of the element makes it resemble a more Proterogyrinus-like tibia, although it lacks the Y-shaped crest observed on the flexor surface of the tibiae of Tulerpeton, Ossinodus and Proterogyrinus

(Holmes, 1984; Lebedev and Coates, 1995; Warren and Turner, 2004). NSM.005.GF.045.320 only slightly resembles tibia type 2 (NSM.014.GF.36.004) (Anderson et al 2015), which is considered Pederpes or Ossinodus-like. The element resembles a form transitional between

Ossinodus and Proterogyrinus, likely an early embolomerous form.

Both tibiae are more concave along the posterior edge than Crassigyrinus.

NSM.014.GF.36.004 tibia type 2 is previously described in Anderson et al. (2015) and was borrowed from the NSM for this study. It is most consistent with a Pederpes-like or

Ossinodus-like whatcheeriid (Anderson et al., 2015).

Table 2.1 summarizes the elements borrowed for this study and their morphotype associations. Expected locomotor behaviour are included in the table, however proxies for

46

locomotor behaviour are based on literature which relies on morphology from features other than stylopod and zeugopod elements (Holmes, 1984; Carroll, 1992a; Lebedev and Coates, 1995;

Lombard and Bolt, 1995; Warren and Turner, 2004; Clack and Finney, 2005)

Table 2.1. Borrowed Blue Beach elements and their attributed taxonomic morphologies and lifestyles. New morphotypes are in bold.

Specimen Element Type Comparable To Locomotor Behaviour

YPM.PU.23545 Humerus Type 1 Whatcheeriid (Ossinodus- Amphibious to like/Pederpes-like) Terrestrial

YPM.PU.20754 Humerus Type 2 Embolomerous Amphibious to Terrestrial

NSM.007.GF.004.733 Ulna Type 1 Tulerpetontid Aquatic

NSM.014.GF.036.002 Radius Type 1 Acanthostegid Aquatic

NSM.007.GF.004.767 Radius Type 1 Acanthostegid Aquatic

NSM.005.GF.045.048- Femur Type 2 Embolomerous Amphibious to A Terrestrial

RM.20.6711 Femur Type 4 Acanthostegid Aquatic

NSM.007.GF.004.630 Femur Type 1 Tulerpetontid Aquatic

NSM.005.GF.045.042 Femur Type 1 Tulerpetontid Aquatic

NSM.005.GF.045.044 Distal Type 5 Whatcheeriid (Ossinodus-like) Amphibious to Femur Terrestrial

NSM.005.GF.045.047 Femur Type 5 Whatcheeriid (Ossinodus-like) Amphibious to Terrestrial

YPM.PU.23550 Femur Type 1 Tulerpetontid Aquatic

YPM.PU.20103 Femur Type 5 Whatcheeriid (revised) Aquatic to Amphibious

NSM.005.GF.045.061 Fibula Type 1 Whatcheeriid (Pederpes-like) Amphibious to Terrestrial

NSM.005.GF.045.063 Fibula Type 1 Whatcheeriid (Whatcheeria- Aquatic like)

NSM.005.GF.045.112 Fibula Type 2 Crassigyrinid Aquatic to Terrestrial?

47

RM.20.5222 Tibia Type 3 Embolomerous Amphibious to Terrestrial

NSM.005.GF.045.320 Tibia Type 3 Embolomerous Amphibious to Terrestrial

Discussion

Initial sampling and descriptions of some of the material at Blue Beach in 2015 exposed a diverse fauna present in the early Carboniferous. The site includes elements similar to multiple

Carboniferous animals such as the embolomerous Proterogyrinus-like and Eoherpeton-like; whatcheeriid Pederpes-like, Ossinodus-like and Whatcheeria-like; and Crassigyrinus-like

(Mansky and Lucas, 2013; Anderson et al., 2015; Lennie et al. In Press). The animals at Blue

Beach are comparable to these taxa which are previously known from disparate temporal ranges

(North America, Britain, and Australia) and cross over the end Devonian mass extinction. Their presence in the Early Carboniferous complicates ideas about a post extinction period of low diversity or faunal turnover between Devonian and Carboniferous periods (Sallan et al., 2010;

Smithson et al., 2012). Whereas some elements are most comparable to a single known taxon

(ex., Tulerpeton-like femur NSM.005.GF.045.042, or Pederpes-like fibula

NSM.005.GF.045.061), others appear to have a suite of features comparable to multiple known

Devonian and Carboniferous tetrapods (ex., YPM.PU.20103, which is comparable to both the

Devonian Tulerpeton (Anderson et al., 2015) and Ossinodus (Warren and Turner, 2004)).

Temporally disparate body forms present at Blue Beach also capture a diverse range of proposed body forms and locomotor behaviour. For example, the tetrapod Crassigyrinus scoticus, known from the Visean of Scotland has been suggested as a secondarily aquatic animal (Carroll, 1992b;

Herbst and Hutchinson, 2019) and Crassigyrinus-like elements (Mansky and Lucas, 2013;

48

Lennie et al. In Press) at Blue Beach could represent a better terrestrially adapted animal.

Animals comparable to Eoherpeton and Proterogyrinus could represent more terrestrial animals, as both of those taxa have cranial morphology associated with more terrestrial rather than aquatic forms (Holmes, 1984; Smithson, 1985). Additionally, elements similar to those of the early tetrapod Pederpes (which exhibits asymmetry in its manus and was proposed by Clack and

Finney, 2005 as a terrestrially adapted animal), add to the diversity of earliest known potentially terrestrial animals at Blue Beach.

Most of the fossils at Blue Beach preserve rugosity and foramina despite exposure to weathering. This may be informative regarding muscle insertion and inferences about other soft tissue interactions, including blood vessel and nerve paths through or along the bones. It is difficult to determine the depth of the pits/foramina in an external analysis, but this topic will be discussed in later chapters. Small pits appear common on anterior edges of femora, while larger, irregularly shaped pits are present within the intertrochanteric fossae, popliteal areas and fibular fossae. The distal elements also have many rugosities. Large elongate pits were preserved in all the distal elements excluding NSM.007.GF.004.733 and NSM.007.GF.004.767. Prominent ridges and/or crests were also observed in all the distal elements. This likely indicates extensive soft tissue interaction with the distal elements. Prominent adductor blades and crests on the femora likely indicate musculature exerting forces along the shafts of the proximal bones. This is unusual compared to modern amphibians and amniotes where most of the musculature interacts with the proximal and distal ends of the femora (Petermann and Sander, 2013; Molnar et al.,

2018).

The three-dimensional preservation of most of the material will enable study of internal structures of these elements. Improvement of modern methods regarding non-destructive

49

sampling and increasing crossover between bone biomechanics and the study of fossil material increases the information we can get from these isolated elements. Despite being isolated and worn, these elements may better reveal how animals were locomoting across the fin-to-limb transition, and when terrestrial locomotion evolved.

Conclusion

The diverse collection of tetrapod material present at Blue Beach ranges from Devonian

Acanthostega-like morphotypes to Carboniferous embolomere and anthracosaur-like forms. The elements described here support previous identifications (Anderson et al., 2015) and contribute descriptions of new material consistent with Devonian Acanthostega-like and Tulerpeton-like elements. A Pederpes-like fibula, and recognition of more Ossinodus-like limb morphotypes supports the presence of multiple Carboniferous whatcheeriid forms, likely with more amphibious habits (Clack and Finney, 2005) and coexisting with Devonian-like animals. The presence of a Crassigyrinus-like fibula is especially intriguing because of previous assertions that Crassigyrinus scoticus is a secondarily aquatic animal (Carroll, 1992; Herbst and

Hutchinson, 2019). The Crassigyrinus-like fibula from Blue Beach may represent the intermediate terrestrial form and contributes to the expansion of the known temporal and geographical range of this group of animals. The Blue Beach limb bones have a range of features comparable to animals from disparate temporal and geographical locations. The morphological features on a single Blue Beach limb bone which are consistent with multiple better known taxa may be indicative of transitional morphologies. Alternatively these may represent a last common ancestor of deep diverging relationships between better known taxa. As we gain comparative material, increased sampling of Tournaisian localities will aid in identification of the isolated

Blue Beach elements. This locality’s diverse overlap in Devonian-like and Carboniferous-like

50

fauna suggest that there was no clear extinction and replacement or faunal turnover between

Devonian and Carboniferous. The range of morphologies of limb elements from the Early

Carboniferous at Blue Beach supports the occurrence of an early diversification of body plans during the fin-to-limb and water-to-land transitions.

Literature Cited

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Carroll, R. L., E. S. Belt, D. L. Dineley, D. Baird, and D. C. McGregor. 1972. International

Geological Congress: Twenty Fourth Session. 1–113 pp.

Carroll, R. L. 1992. The Primary Radiation of Terrestrial Vertebrates. Annual Review of Earth

and Planetary Sciences 20:45–84.

Clack, J. A. 2001. Eucritta melanolimnetes from the Early Carboniferous of Scotland, a stem

tetrapod showing a mosaic of characteristics. Transactions of the Royal Society of

Edinburgh: Earth Sciences 92:75–95.

Clack, J. A. 2002. An early tetrapod from “Romer’s Gap”. Nature 418:72–6.

Clack, J. A., C. E. Bennett, D. K. Carpenter, S. J. Davies, N. C. Fraser, T. I. Kearsey, J. E. A.

Marshall, D. Millward, B. K. A. Otoo, E. J. Reeves, A. J. Ross, M. Ruta, K. Z. Smithson, T.

R. Smithson, and S. A. Walsh. 2017. Phylogenetic and environmental context of a

Tournaisian tetrapod fauna. Nature Ecology & Evolution 1.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

51

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Clack, J. A., R. L. Carroll, and H. Heatwole. 2000. Early Carboniferous tetrapods; pp. 1030–

1043 in Amphibian Biology. vol. 4 Palaeont.

Clack, J. A., L. B. Porro, and C. E. Bennett. 2018. A Crassigyrinus -like jaw from the

Tournaisian (Early Mississippian) of Scotland. Earth and Environmental Science

Transactions of the Royal Society of Edinburgh 108:37–46.

Clack, J. A., C. E. Bennett, S. J. Davies, A. C. Scott, J. E. Sherwin, and T. R. Smithson. 2019. A

Tournaisian (earliest Carboniferous) conglomerate-preserved non-marine faunal assemblage

and its environmental and sedimentological context. PeerJ 2019:e5972.

Coates, M. I. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: Postcranial anatomy,

basal tetrapod interrelationships and patterns of skeletal evolution. Transactions of the

Royal Society of Edinburgh, Earth Sciences 87:363–421.

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

Hall, B. K. 2007. Fins into Limbs: Evolution, Development, and Transformation. University of

Chicago Press, 433 pp.

Herbst, E. C., and J. R. Hutchinson. 2019. New insights into the morphology of the

Carboniferous tetrapod Crassigyrinus scoticus from computed tomography. Earth and

Environmental Science Transactions of the Royal Society of Edinburgh 109:157–175.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

52

Sciences 306:431–524.

Jarvik, E. 1996. The Devonian tetrapod Ichthyostega. Lethaia 29:76–76.

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Lombard, R. E., and J. R. Bolt. 1995. New Primitive Tetrapod Whatcheeria deltae From Lower

Carboniferous Iowa. Palaeontology 38.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, A. T., D. C. McGregor, and J. Utting. 1993. Stratigraphic significance of Upper

Devonian and Lower Carboniferous miospores from the type area of the Horton Group,

Nova Scotia. Canadian Journal of Earth Sciences 30:1091–1098.

Martel, T. A., and M. R. Gibling. 1996. Stratigraphy and tectonic history of the Upper Devonian

to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology 32:13–38.

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2018. Evolution of Hindlimb Muscle

Anatomy Across the Tetrapod Water-to-Land Transition, Including Comparisons with

Forelimb Anatomy. Anatomical Record. doi: 10.1002/ar.23997.

Otoo, B. K. A., J. A. Clack, T. R. Smithson, C. E. Bennett, T. I. Kearsey, and M. I. Coates. 2019.

A fish and tetrapod fauna from Romer’s Gap preserved in Scottish Tournaisian floodplain

deposits. Palaeontology 62:225–253.

Panchen, A. L. 1985. On the Amphibian Crassigyrinus scoticus Watson from the Carboniferous

53

of Scotland. Philosophical Transactions of the Royal Society of London. Series B 309:505–

568.

Panchen, A. L., and T. R. Smithson. 1990. The pelvic girdle and hind limb of Crassigyrinus

scoticus (Lydekker) from the Scottish Carboniferous and the origin of the tetrapod pelvic

skeleton. Transactions of the Royal Society of Edinburgh: Earth Sciences 81:31–44.

Petermann, H., and M. Sander. 2013. Histological evidence for muscle insertion in extant

amniote femora: Implications for muscle reconstruction in fossils. Journal of Anatomy

222:419–436.

Romer, A. S. 1957. The appendicular skeleton of the Permian embolomerous amphibian

Archeria. Contributions from the Museum of Paleontology, University of Michigan 13:103–

159.

S. J. Godfrey. 1989. The Postcranial Skeletal Anatomy of the Carboniferous Tetrapod

Greererpeton burkemorani Romer, 1969. Philosophical Transactions of the Royal Society

B: Biological Sciences 323:75–133.

Sallan, L. C., M. I. Coates, J. Clack, L. C. Saltan, and M. I. Coates. 2010. End-Devonian

extinction and a bottleneck in the early evolution of modern jawed End-Devonian extinction

and a bottleneck in the early evolution of modern jawed vertebrates. Proceedings of the

National Academy of Sciences of the United States of America 107:10131–10135.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Smithson, T. R., S. P. Wood, J. E. A. Marshall, and J. A. Clack. 2012. Earliest Carboniferous

54

tetrapod and arthropod faunas from Scotland populate Romer’s Gap. Proceedings of the

National Academy of Sciences of the United States of America 109:4532–7.

Smithson, T. R., and J. A. Clack. 2013. Tetrapod appendicular skeletal elements from the Early

Carboniferous of Scotland. Comptes Rendus - Palevol 12:405–417.

Smithson, T. R., and J. A. Clack. 2018. A new tetrapod from Romer’s Gap reveals an early

adaptation for walking. Earth and Environmental Science Transactions of the Royal Society

of Edinburgh 108:89–97.

Sues, H.-D., R. W. Hook, and P. E. Olsen. 2013. Donald Baird and his discoveries of

Carboniferous and early Mesozoic vertebrates in Nova Scotia. Atlantic Geology 49:90.

Tibert, N. E., and D. B. Scott. 1999. Ostracodes and Agglutinated Foraminifera as Indicators of

Paleoenvironmental Change in an Early Carboniferous Brackish Bay, Atlantic Canada.

PALAIOS 14:246.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

Warren, A. 2007. New data on Ossinodus pueri, a stem tetrapod from the Early Carboniferous of

Australia. Journal of Vertebrate Paleontology 27:850–862.

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Waters, C. N., R. A. Waters, W. J. Barclay, and J. R. Davies. 2009. A Lithostratigraphical

Framework for the Carboniferous Successions of Northern Great Britain (Onshore). 174 pp.

55

Chapter 3: Musculature comparisons using extant taxa Introduction Reconstructing soft tissue in fossil taxa can be useful to learn about the biomechanics and locomotor behaviours of early tetrapods. The fin-to-limb transition, which was previously obscured by a gap in the fossil record, is rapidly being filled by fragmentary and complete, articulated, material, and much remains to be learned about the animals that left these fossils.

Locomotor behaviours were previously inferred from external morphological proxies, such as the presence of digited limbs (Barrell et al., 1920; Inger, 1957). However, with evidence of internal gills in the digited tetrapod Acanthostega (Coates and Clack, 1991), new proxies for terrestrial locomotion are needed. Many paleontologists make observations about limb proportion, and digit number (Coates 1996; Coates & Clack 1995; Warren & Turner 2004; Clack & Finney

2005; Coates & Clack 1990) to suggest aquatic or more terrestrial lifestyles.

Whereas these proxies for locomotor behaviour have their appeal, they cannot be used for lifestyle inference in Blue Beach tetrapods, which are represented by single elements found in isolation from each other. The Extant Phylogenetic Bracket (EPB; Bryant and Russell, 1992;

Witmer, 1995) is a method by which the confidence of inferences of soft tissue can be examined

(Hutchinson, 2001b, 2001a, 2002; Diogo et al., 2008; Abdala and Diogo, 2010; Sanchez et al.,

2010; Pierce et al., 2012; Diogo et al., 2015, 2016; Molnar et al., 2017, 2018). Osteological correlates of soft tissue are compared between outgroup and extant sister taxa of the fossil of interest. Using these bracket taxa inferences of identity and functionality of osteological correlates are made with varying degrees of confidence (Bryant and Russell, 1992; Witmer,

1995). The EPB is less useful for testing inferences when bracketing taxa have deep divergences with the fossil taxa of interest (Cloutier and Ahlberg, 1996; Bishop, 2014), such as is the case for stem tetrapods. Bryant and Seymour (1990) suggest that to infer any soft tissue attachments that

56

Figure 3.1 Cladogram showing nested groups more closely related to each other than the taxa to the left. For example, birds and crocodiles form a sister group to lepidosaurs, and are more closely related to each other than they are to lepidosaurs. Lepidosaurs, crocodiles, and birds form a sister groups with mammals, and are more closely related to each other than they are to mammals. And so on. The silhouettes illustrate there is a great deal of variation in body plans, and all major extant groups have some taxa that have undergone a secondary transition to an aquatic or amphibious lifestyle. have dubious or no osteological correlates, extremely similar taxa are required for comparison.

For early tetrapods no such similar taxa exist. Stem tetrapods are bracketed by lungfish on one side and all living amphibians and amniotes on the other (Figure 3.1). The lungfish of the

Paleozoic diverged from the tetrapod lineage over 400mya, leading to a large amount of divergent evolution. Cloutier and Ahlberg (1996) point out that poor understanding of some morphological characters used in cladistic analyses makes it difficult to distinguish homoplasy

57

from homology over these evolutionary distances. Additionally, modern lungfish are very derived with respect to the lungfish that diverged from the tetrapod lineage over 400mya, having heavily reduced fins as one unusual feature (Schoch, 2014), further confounding comparative anatomy. Similarly, within the descendant bracket (crown group tetrapods) is a diverse array of animals that includes all living amphibians and amniotes that diverged from each other approximately 370mya. Hundreds of millions of years between the divergences of stem lungfish and stem tetrapods, and subsequent divergent evolution of crown lungfish and tetrapods leads to neither bracket group having limbs with morphological features similar to those seen in the taxa of interest. This lack of morphological similarity to early tetrapods of both bracketing taxa leaves the EPB problematic for reconstructing soft tissues in this case. For this reason, this study will not use the EPB.

Simple observations of size, location, and directionality of external osteological correlates can still provide useful information about how and where forces acted on the bone

(Petermann and Sander, 2013; Bishop, 2014), and thus locomotor behaviour. I will compare external morphological features of Blue Beach femora to those of extant amphibians and amniotes that have aquatic, amphibious, and terrestrial locomotor patterns.

Blue Beach has a great number of isolated fossil limb bones that are three-dimensionally preserved. While the specimens were disarticulated and transported before deposition and are currently eroded out by tidal action (Mansky and Lucas, 2013), surface features like foramina are clearly visible. The greatest number of these elements are femora, so this study will focus on them. Morphotypes of Blue Beach material have previously been attributed to a range of

Devonian to Carboniferous material (Anderson et al., 2015). Devonian taxa are represented by acanthostegid and tulerpetontid elements, animals that lived aquatic lifestyles based on evidence

58

of internal gills (Coates and Clack, 1991) and depositional environment (Lebedev and Coates,

1995). Carboniferous taxa are represented by whatcheeriid and embolomerous elements, which may represent animals more capable of terrestrial locomotion based on orientation of the foot

(Clack and Finney, 2005) and more ‘reptilopmorph’ skull features (Holmes, 1984; Smithson,

1985). The site is early-mid to mid Tournaisian (CM and PC palynozones; Utting et al., 1989;

Martel and Gibling, 1996), and possibly represents the earliest known Carboniferous locality.

However, review of the Ballagan formation of Scotland suggests sites in Scotland extend below the CM palynozone as well (Waters et al., 2009; Bennett et al., 2017).

The aims of this study are to 1) identify muscle scarring locations and orientations associated with extant aquatic and terrestrial taxa, 2) identify trends in the shape or direction of major bone features (such as trochanters) that may indicate aquatic or terrestrial lifestyle, and 3) identify locations and patterns of scarring on Blue Beach limb bones to see if they are comparable to the extant taxa examined.

Institutional abbreviations: Blue Beach Fossil Museum, Hantsport, Nova Scotia, Canada

(BBFM); Musée d’Histoire Naturelle de Miguasha, Québec, Canada (MHNM); Museum of

Vertebrate Zoology, University of California, Berkeley, USA (MVZ); Nova Scotia Museum of

Natural History, Halifax, Canada (NSM); Redpath Museum, McGill University, Montreal,

Canada (RM); University of Michigan Museum of Zoology, Ann Arbor, Michigan, USA

(UMMZ); Yale Peabody Museum of Natural History, New Haven, Connecticut, USA (YPM);

Museum für Naturkunde Berlin, Germany (ZMB).

Materials and Methods Hindlimb elements, particularly femora, will be the primary subjects of this analysis as they were the most common fossil bone in the sample and are used for aquatic locomotion

59

(kicking) and terrestrial propulsion (crawling or walking). Both forms of locomotion involve different limb orientations, behind the body for kicking, and lateral and ventral to the body for crawling.

Necturus lives in sandy streams and was used to represent a fully aquatic animal (Neill,

1963; Craig et al., 2016) with musculature specialised for aquatic locomotion. Cynops, a genus of newt, represents an amphibious animal which spends a great deal of its adult life in a terrestrial environment, but many of the species in the genus require an aquatic environment to reproduce and mature (Sparreboom and Ota, 1995). Uromastyx is a medium sized lepidosaur (the biggest of the four animals used for comparison) and is known to live in dry desert environments where it inhabits burrows (Cunningham, 2000). Eublepharis, the leopard gecko, is another terrestrial quadruped that lives in arid terrestrial environments (Fuller et al., 2011). One femur of

Felis (domestic cat, and terrestrial carnivore) was also used as a comparative bone sample in this study.

I examined extant and extinct femora using micro computed tomography (microCT). The skeletons of Cryptobranchus (FMNH UF10881), Amblyrhynchus (FMNH UF41558) and

Ornithorhynchus (MVZ MVZ32885) were downloaded from Morphosource

(morphosource.org), an online repository of CT scans. The remaining taxa were obtained from previous projects, and desiccation from freezing and thawing occurred prior to the current study.

These animals were microCT scanned at the University of Calgary to enable visualization of the bones. In the fossil scans I observed depth and orientation of pit structures on the bone surface. I also virtually dissected the extant animals via segmentation of musculature using diffusible iodine-based contrast enhanced computed microtomography (diceCT; Gignac et al., 2016).

DiceCT scans were performed to determine the differentiation and location of musculature and

60

its interaction with bone in all three locomotory behaviours. The amphibian Cynops and amniotes Eublepharus and Uromastyx were previously fixed in neutral buffered formalin and stored in 70% ethanol. Necturus was stored in Wards Solution.

Before preparation for diceCT the animals were scanned to obtain a bone only sample.

They were then placed under running tap water overnight (up to 24 hours). I removed the limbs of Necturus and placed them in a solution of 11.24% Lugol’s iodine (Gignac et al., 2016) on a shaker (Stoval Life Sciences, Inc) for two days prior to scanning. Agitation from the shaker facilitates faster and even diffusion of Lugol’s iodine across the tissues (Gignac et al., 2016). A whole Cynops specimen was placed in Lugol’s iodine and continuously agitated for two weeks prior to scanning. The amniotes Eublepharis and Uromastyx were skinned from the neck to the anterior end of the tail (including all four limbs) before staining to enhance diffusion of iodine into the limb musculature (Gignac et al., 2016). After skinning, the amniotes were immersed in

Lugol’s iodine and continuously agitated for eight (Eublepharis) and fourteen (Uromastyx) days.

All of the animals were scanned in a Skyscan 1173 (Bruker, Belgium) at the University of

Calgary with settings between 80-120kV and 70-85µA with a 9.6-30µm voxel size. When necessary, diceCT image slices were subjected to a contrast-limited adaptive histogram equalization (CLAHE) filter in dragonfly (ORS, Montréal) to enhance contrast between muscle and other tissues (Gignac and Watanabe, 2016). The resulting diceCT scan images were volumized using Amira version 5.3.3 (Visage, Inc) to segment individual muscles in a virtual dissection.

After each animal was scanned with Lugol’s iodine they were rinsed using tapwater and replaced in ethanol. The first Cynops bone scan did not have well preserved limb bones so an additional animal was scanned. The original Uromastyx femora scanned had unossified proximal

61

and distal ends, and an unusual knob in the middle of the proximal articulating surface, and an additional specimen was found to have the same morphology. Isolated bone scans were opened in Dragonfly 4.0 (ORS, Montréal) for analysis. The smallest voxel size for bone scans was

9.6µm. Necturus and Uromastyx hindlimbs were additionally manually dissected and photographed using a Nikon BM-6 camera and Nikon SMZ1500 lens attachment.

Fossil femora were prepared by Sonja Wood and KL and scanned in the Xradia microCT scanner (Zeiss) at the McCaig Institute for Bone and Joint Health at the University of Calgary.

All fossil femora were scanned at a magnification of 0.4X, and scan parameters of 140kV 10

Watts and 71.7µA. Voxel sizes ranged between 14.3 and 30.2µm. Descriptions of the fossil femora can be found in Chapter Two.

Results Unfortunately, diceCT scans of the musculature were not capable of visualizing tendon, so I was unable to observe locations of tendinous muscle insertions (Appendix A). After a manual dissection of Necturus and Uromastyx, it became clear that many of the muscles attaching to the shaft of the bone do not leave prominent osteological correlates comparable to the ridges and rugosities observed along the shafts of the Blue Beach femora. The manual dissection of Necturus exposed that much of the musculature attaches to the unossified cartilaginous caps at the proximal and distal ends of the femora rather than the bones themselves.

As cartilage does not preserve in most fossil material, potential inferences based on comparison between muscle attachment sites in Necturus and Blue Beach femora are not possible.

62

Figure 3.2. Progression of sampled aquatic to amphibious to terrestrial femora sampled. µCT images ofA, femur of Necturus (aquatic). B, femur of Cynops (amphibious). C, femur of Uromastyx (terrestrial). D, femur of Eublepharis (terrestrial). Scale bars represent 1mm.

Large Osteological Features

Necturus (Figure 3.2 A) and Cryptobranchus femora both have broad flat proximal heads

which are rotated (not quite 90 degrees) to the distal end. The articulating surfaces of Necturus

femora are completely cartilaginous and the end of an isolated bone in proximal view is a hollow

tube. The femora of both taxa have a tubular trochanter that is discontinuous and separates from

the diaphyseal shaft and originates approximately one third of the way down the bone. The

proximal end of the trochanter does not extend as far as the proximal head of the femur in

Necturus or Cryptobranchus. The trochanters in both animals diverge from the main shaft at an

approximate 30-45 degree angle which creates a distinct V-shape. Both animals have a small

bony protrusion where the trochanter attaches to the shaft of the bone. The m.

puboischiofemoralis externus (PIFE) attaches to the trochanter in Salamandra (Molnar et al.,

2018), however exact size and place of attachment as well as direction of muscle force cannot be

determined from examination of the bone. Both shafts are narrow, smooth, and cylindrical before

the distal end becomes broad and flat. The bones are smooth and lack rugosity. The femur of the

63

more terrestrial amphibian Cynops (Figure 3.2 B) is different than those of Necturus and

Cryptobranchus. The proximal end is more completely ossified, and the trochanter is less tube- like and more continuous with the main shaft and proximal end of the femur. The proximal head is rotated less than 90 degrees relative to the distal end of the bone. The Cynops femur is also much more rugose than either of the fully aquatic amphibians, and most similar to the early tetrapod bones (Figure 3.3). The proximal head of the femur appears to be ossified bone and is more complicated in outline than the simple oval of the aquatic amphibians. The femur of

Cynops has a well-defined trochlear groove lined by two trochlear crests that run from the head to proximal of the trochanter. The trochanter of Cynops is more continuous with the proximal head than in the aquatic amphibians. The trochanter meets the shaft in a U-shape rather than a V- shape. There is a raised ridge running proximodistally down the flexor surface of the trochanter.

The ridge runs along the ventroanterior edge of the bone before bifurcating distally.

One part of the bifurcation extends anteriorly and the other towards the midline. Proximal to the bifurcation the ridge is consistent with the location of the PIFE (Molnar et al., 2018). The PIFE functions in newts to retract the femur and support the body (Walthall and Ashley-Ross, 2006).

Distal to the bifurcation the muscle the ridge may correlate to is unclear. A less well-developed posterior ridge connects with the anterior ridge at approximately the midshaft. The location of this ridge is consistent with the attachment site of the m. puboischiofemoralis internus (PIFI) in the salamander femur (Molnar et al., 2018). The PIFI is a femoral protractor and elevator

(Walthall and Ashley-Ross, 2006). Russell and Bauer (2008) and Petermann and Sander (2013) observe the PIFE and PIFI have a combination of fleshy and tendinous attachments across multiple sites in both Iguana (Russell and Bauer 2008) and Alligator (Petermann and Sander,

2013). Therefore, observable tendinous correlates are not surprising in these regions. It is

64

interesting that the ridges are most prominent in Cynops and the Blue Beach femora. Aquatic amphibians and amphibious amniotes lack prominent ridges along the midshaft.

Cynops forages both on land and in water (Sparreboom and Ota, 1995) unlike Necturus and Cryptobranchus. The rounded proximal head, rugosity associated with protractor and retractor muscles, and U-shape as opposed to V-shape between trochanter and femoral head, might be indicative of a more amphibious or terrestrial lifestyle.

Eublepharis (Figure 3.2 D) has an elongate femur with a distinct neck, head, and trochanter. The head and trochanter are connected by a U-shaped region of finished bone. The trochanter originates within the proximal quarter of the bone and extends proximally almost as far as the femoral head. The proximal and distal articulating surfaces of the Eublepharis femur are the most completely ossified of the four extant animals examined. The shaft is smooth and lacks rugosity. Any rugosity present on the bone is confined to the distal and proximal neck and trochanter.

Uromastyx (Figure 3.2 C) was the largest and most terrestrial of the animals studied. In contrast to Cynops the proximal head of the femur is less ossified, covered instead by a large cartilaginous cap. I scanned an additional Uromastyx to assess if this was an unusual specimen and it had the same features. A scan of the skull revealed well ossified elements except for the contacts between the frontal, parietal, and postorbital at the posterior of the orbit suggesting a subadult animal; however, as observed in Chapter Five, the trabeculae in this femur are highly aligned. During long bone formation remodelling of trabecular orientation is known to occur in post juvenile animals after changes in bone volume fraction (Tanck et al., 2001). Uromastyx is additionally unusual because it does not have an obvious rounded head or trochanter. The head appears moderately oval and is angled towards the pelvis. There is a small neck which leads into

65

Figure 3.3. Comparison of topological similarities between the femora of Cynops (A,C) and Blue Beach YPM.PU.20103 (B,D). Yellow appears to resemble an intertrochanteric fossa, blue the popliteal area, and purple a fibular fossa. Red represents well developed ridges

66

a dorsoventrally flattened, ventral continuation of the shaft of the bone. Distal to this, the shaft is cylindrical with a distinct nutrient foramen near the midshaft. The distal end is slightly broadened with a small upward edge on the ventral surface. The distal articulating surface is more ossified than the proximal surface.

The only pits observed in the microCT scans of the extant animals examined (some at voxel sizes of 9.6µm) were those of nutrient foramina located near the midshaft of the elements.

This illustrates that pits associated with muscle insertion in modern amphibians and amniotes are too small to be viewed without histology or scanning technology at the sub micron level.

Extensive small (0.1-0.3mm) pits on the anterior, posterior, and dorsal surfaces radiate away from the midshaft on the femora NSM.007.GF.004.630, NSM.005.GF.045.048-A, RM.20.6711, and YPM.PU.23550. YPM.PU.20103 lacked such small pits, and the other femora were too incomplete to determine if pitting or lack of pitting was because of wear or a biological feature.

These small pits on the Blue Beach femora are consistent in size with those observed and described as attachments sites for muscle fascicles in the humerus of Ossinodus (Bishop, 2014).

Larger rugose pits are present within the intertrochanteric fossae, popliteal areas, and fibular fossae of the fossil femora examined.

No pits similar (in location, size, or number) to those covering extensive surface areas on

Blue Beach fossil bones were present in the extant femora, despite the higher resolution of the extant femora scans. Therefore, I was unable to determine if presence and/or location of pits on the fossil limb bone are indicative of any kind of lifestyle. I was also unable to differentiate regionalization of the small pits, perhaps indicating that they are associated with a single large muscle mass. However, scans of the fossil femora revealed that both large rugose (Figure 3.4) and numerous small pits (Figure 3.5) extend through the cortical bone into the trabecular cavity.

67

Discussion Gross patterns in bone morphology in representative extant taxa and comparison to the Blue Beach material. Interestingly, if Uromastyx had not been included in this analysis an attractive sequence from Necturus (aquatic; Figure 3.2 A) to Cynops (amphibious; Figure 3.2 B), to Eublepharis

(terrestrial, Figure 3.2 D) displays a progression from aquatic to terrestrial femora. The proximal ends are rounded out, and distal ends of the bones progress from flattened in aquatic animals to cylindrical in more terrestrial animals. Both proximal and distal ends of the aquatic femora are less ossified than the proximal and distal ends of the more terrestrial femora. Additionally, the trochanters from Necturus to Eublepharis become less tube-like, more proximally located, and attached via a U-shaped (rather than V-shaped) ridge of finished bone. However, the femur of

Uromastyx (Figure 3.2 C), a terrestrial animal, interrupts this trend (Figure 3.2). It lacks a distinct trochanter, has less ossified cortical bone (finished bone) on the proximal end, and the proximal end is flattened as opposed to rounded.

The Blue Beach femora are cylindrical at the midshaft. The proximal and distal ends have varying morphologies; from flattened (NSM.005.GF.045.048-A), similar to that of Necturus or

Uromastyx, to bulbous (NSM.007.GF.004.630). A flattened proximal femur does not convey information regarding terrestrial potential in the femora of extant animals examined. Therefore, it cannot be used to infer such information in the fossil femora. The bulbous proximal end of

NSM.007.GF.004.630 does not resemble a femoral head. There is no neck and the bulbous morphology is continuous with the shaft of the bone. The fossil femora are not as elongate or gracile as either the amphibious Cynops or terrestrial Eublepharis and Uromastyx. However, similar to Uromastyx and Necturus, the Blue Beach femora have proximal and distal articulating

68

Figure 3.4. Tetrapod femur NSM.007.GF.004.630. A, green arrows indicate where large rugose pits extend through the cortical bone. B, same pits in A visible in transaxial sections crossing through the cortical layer. C, same pits viewed in longitudinal sections crossing the cortical layer.

69

70

Figure 3.5. Femur NSM.007.GF.004.630. Green arrows are pointing to small “pits” found across blue beach taxa, similar in size to those described by Bishop (2014). B, same pits in A visible in transaxial sections crossing through the cortical layer. C, same pits viewed in longitudinal sections crossing the cortical layer. surface that lack finished bone. The fossil elements are not hollow cylinders like Necturus. The proximal articulating surfaces of the Blue Beach femora more closely resemble Uromastyx, a convex structure with exposed trabecular bone. The trochanters of the Blue Beach femora are located within prominent adductor blades spanning the middle third of the bone. This ubiquitous adductor blade in the fossil femora is intriguing because adductor muscles in modern amniotes generally attach directly to the bone and do not leave any macroscopic trace (Bryant and

Seymour, 1990; Petermann and Sander, 2013). In the extant animals studied here the terrestrial and even amphibious taxa have trochanters at the proximal ends of the bone and the aquatic taxa have more distally located (but still proximal) trochanters. In contrast, the fossil femora have trochanters within the adductor blade much closer the midshaft. It appears surface features of early tetrapod limb bones previously used to infer soft tissue are different enough in early tetrapods to make them incomparable with extant tetrapods.

Cynops was an interesting specimen to examine. There is no published literature on the hindlimb morphology and anatomy of the animal that I could find. The scans of Cynops reveal rugose areas similar to those of the intertrochanteric fossa (Figure 3.3, yellow), fibular fossa

(Figure 3.3, purple), and popliteal areas (Figure 3.3 purple) of the femora of early tetrapod

YPM.PU.20103 (Figure 3.3). In Cynops the rugose areas are depressed and demarcated by well- developed ridges (Figure 3.3, red), associated with tendinous muscle attachment sites consistent with the PIFE, and PIFI (Russell and Bauer, 2008; Petermann and Sander, 2013; Molnar et al.,

2018) muscles. The PIFE is involved in retracting the femur and supporting the body, whereas the PIFI opposes the action of the PIFE, by protracting and elevating the femur (Walthall and

71

Ashley-Ross, 2006). Well-developed rugosities in the areas of these muscle attachment sites are suggestive that the femur is used for a range of crawling motions rather lateral paddling. In the early tetrapods (here exemplified by whatcheeriid YPM.PU.20103) the intertrochanteric fossae, fibular fossae and popliteal areas are depressed and highly rugose regions with less developed ridges (excluding the adductor crest; Figure 3.3). The presence of less developed ridges suggests musculature was acting to protract and retract the fossil femora, indicative of movement more complex than lateral paddling.

Muscle maps of the femur of Salamandra (Molnar et al. 2018) show no muscles attaching to the bone within the region of the intertrochanteric fossa, and in mammalian bone, the popliteal fossa is not an area of muscle attachment. Additionally, the observation that the pits making up the rugose areas within these fossae in early tetrapod femora are 1) too big to be tendinous insertion fiber pits, and 2) penetrate through the entire cortical bone and into the trabecular cavity, indicates that many rugose areas are not associated with muscle attachment.

Other muscle correlates may be evident, and features such as trochanters and ridges can still be examined. However, muscle reconstructions based only on macroscopic osteological correlates should be viewed as incomplete pictures, and should not be the only method used to construct and test hypotheses requiring complete reconstructions of soft tissue systems.

Extensive pitting of early tetrapod femora Jones and Boyde (1974) examined Sharpey’s fibers in a small number of animals and found the average size for the insertion was 10µm, and (Sharpey’s) fiber insertion size did not vary with size of the animal. This is consistent with the size of muscle fiber insertions figured in

Sanchez et al. (2013: figure 8 a). Given ambiguous references to Sharpey’s fibers in the literature

72

(Hutchinson and Gatesy, 2000; Hutchinson, 2001a, 2002), it must be pointed out that remnants of muscle insertion fiber pits are not visible at the macroscopic level.

Bishop et al. (2014) attempted to differentiate patterns of rugosity observed on the humerus of Ossinodus. He describes pits, which do not extend through the cortical layer and are lined with finished bone. The pits have 100µm to 300µm diameters and are up to 3mm long.

They are attributed to muscle fascicle insertion pits rather than (Sharpey’s) muscle fiber insertion pits. This description also matches extensive small pitting on the Blue Beach femora, only the pits extend through the cortical bone and into the trabecular space in the Blue Beach material

(Figure 3.5). Additionally, I did not observe any similar pitting on any of the femora of the modern animals used in this study. Therefore, it is likely that previous analyses that attributed pitting and orientation of pitting to muscle attachment fibre sites need to be re-examined.

The pits extending through the cortical bone in Blue Beach femora could be associated with developmental stage or a unique feature of the bones of early tetrapods which does not persist in modern animals. Future research should focus on understanding how tissues are interacting with the bone, and what the various interactions may be used to infer in early tetrapods.

Conclusions Whereas this study did not accomplish its original goal of inferring locomotion based on reconstructions of musculature differentiation and orientation in fossils, it illustrated the many pitfalls of reconstructing accurate musculature from macroscopic observations in fossils of early tetrapods. Musculature reconstructions alone should not be used to test biomechanical hypotheses regarding early tetrapod locomotory capabilities.

73

External morphological comparison of extant amphibians and amniotes to Blue Beach femora indicate that the fossil tetrapods may have had musculature beneficial for a crawling motion. However, the ridges associated with the crawling muscles of Cynops are not as well developed in the Blue Beach femora. This suggests that musculature was not pulling as forcefully to promote extensive cortical bone remodelling, perhaps because the animals were crawling in an aquatic rather than terrestrial environment. Reconstructing musculature in early tetrapods is difficult and confusing. There is a lack of consistent interpretation of muscle attachment sites in addition to confounding osteological correlates known from the literature.

Further extensive studies on a more diverse array of amphibians and amniotes may illuminate possible osteological correlates associated with some muscles. However, the complete reconstruction of early tetrapod musculature remains an unlikely outcome. Future efforts must focus on defining and understanding the mechanisms responsible for the main osteological features associated with muscle attachment: rugosity (including pits and striations), scars, and prominences. Critical to future investigations is the incorporation of data from internal features of limb bones. Methods that use internal data can provide testable ways to infer locomotor capabilities of early tetrapods. These methods will be discussed further in later chapters.

With the new information regarding muscle pits provided in this paper, and their inconsistency with definitions of Sharpey’s fibres or muscle fiber insertion pits, researchers must re-evaluate how muscles were attaching to the bones of early tetrapods, and consider that they may not have attached in a manner similar to extant tetrapods. The pits are not consistent with large foramina which extend through the cortical layer of early tetrapod limb bones and which become reduced as physiological activities require greater amounts of oxygen and other nutrients so that the size of the foramina become a limiting factor (Lebedev and Coates 1995).They are

74

also inconsistent with the size of known muscle insertion fibres (Sharpey’s fibres). Sanchez et al.

(2013) observed pits in the humerus of Acanthostega more consistent with those described in this chapter, however in the 2013 study it was noted that the vascular network the pitting was attributed to ran parallel to the surface of the bone.

Literature Cited Abdala, V., and R. Diogo. 2010. Comparative anatomy, homologies and evolution of the pectoral

and forelimb musculature of tetrapods with special attention to extant limbed amphibians

and reptiles. Journal of Anatomy 217:536–573.

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Barrell, J., E. Huntington, R. S. Lull, C. Schuchert, and L. L. Woodruff. 1920. The Evolution of

the Earth and Its Inhabitants; a Series of Lectures Delivered before the Yale Chapter of the

Sigma Xi during the Academic Year 1916-1917, 3rd ed. Yale University Press, New Haven,

1–235 pp.

Bennett, C. E., A. S. Howard, S. J. Davies, T. I. Kearsey, D. Millward, P. J. Brand, M. A. E.

Browne, E. J. Reeves, and J. E. A. Marshall. 2017. Ichnofauna record cryptic marine

incursions onto a coastal floodplain at a key Lower Mississippian tetrapod site.

Palaeogeography, Palaeoclimatology, Palaeoecology 468:287–300.

Bishop, P. J. 2014. The humerus of Ossinodus pueri , a stem tetrapod from the Carboniferous of

Gondwana, and the early evolution of the tetrapod forelimb. Alcheringa: An Australasian

Journal of Palaeontology 38:209–238.

Bryant, H. N., and K. L. Seymour. 1990. Observations and comments on the reliability of muscle

75

reconstruction in fossil vertebrates. Journal of Morphology 206:109–117.

Bryant, H. N., and A. P. Russell. 1992. The role of phylogenetic analysis in the inference of

unpreserved attributes of extinct taxa. Philosophical Transactions - Royal Society of

London, B 337:405–418.

Carrano, M. T., and J. R. Hutchinson. 2002. Pelvic and hindlimb musculature of Tyrannosaurus

rex (Dinosauria: Theropoda). Journal of Morphology 253:207–228.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Cloutier, R., and P. E. Ahlberg. 1996. Morphology, Characters, and the Interrelationships of

Basal Sarcopterygians. 445–479 pp.

Coates, M. I. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: Postcranial anatomy,

basal tetrapod interrelationships and patterns of skeletal evolution. Transactions of the

Royal Society of Edinburgh, Earth Sciences 87:363–421.

Coates, M. I., and J. A. Clack. 1990. Polydactyly in the earliest known tetrapod limbs. Nature

347:66–69.

Coates, M. I., and J. a. Clack. 1991. Fish-like gills and breathing in the earliest known tetrapod.

Nature 352:234–236.

Coates, M. I., and J. A. Clack. 1995. Romer’s gap: tetrapod origins and terrestriality. Bulletin Du

Muséum National d’Histoire Naturelle 4:373–388.

Craig, J. M., D. A. Mifsud, A. S. Briggs, J. Boase, and G. Kennedy. 2016. Mudpuppy (Necturus

maculosus maculosus) spatial distribution, breeding water depth, and use of artificial

76

spawning habitat in the detroit river. Herpetological Conservation and Biology 10:926–934.

Cunningham, P. 2000. Daily activity pattern and diet of a population of the spinytailed lizard,

Uromastyx aegyptius microlepis, during summer in the United Arab Emirates. Zoology in

the Middle East 21:37–46.

Diogo, R., V. Abdala, N. Lonergan, and B. A. Wood. 2008. From fish to modern humans -

Comparative anatomy, homologies and evolution of the head and neck musculature. Journal

of Anatomy 213:391–424.

Diogo, R., P. Johnston, J. L. Molnar, and B. Esteve-Altava. 2016. Characteristic tetrapod

musculoskeletal limb phenotype emerged more than 400 MYA in basal lobe-finned fishes.

Scientific Reports 6:37592.

Diogo, R., S. Walsh, C. Smith, J. M. Ziermann, and V. Abdala. 2015. Towards the resolution of

a long-standing evolutionary question: Muscle identity and attachments are mainly related

to topological position and not to primordium or homeotic identity of digits. Journal of

Anatomy 226:523–529.

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

Fuller, P. O., T. E. Higham, and A. J. Clark. 2011. Posture, speed, and habitat structure: Three-

dimensional hindlimb kinematics of two species of padless geckos. Zoology 114:104–112.

Gignac, P. M., and A. Watanabe. 2016. Get Your Brights Brighter & Darks Darker: CLAHE +

diceCT – DiceCT. Available at https://dicect.com/2016/08/11/clahe/. Accessed August 26,

2019.

77

Gignac, P. M., N. J. Kley, J. A. Clarke, M. W. Colbert, A. C. Morhardt, D. Cerio, I. N. Cost, P.

G. Cox, J. D. Daza, C. M. Early, M. S. Echols, R. M. Henkelman, A. N. Herdina, C. M.

Holliday, Z. Li, K. Mahlow, S. Merchant, J. Müller, C. P. Orsbon, D. J. Paluh, M. L. Thies,

H. P. Tsai, and L. M. Witmer. 2016. Diffusible iodine-based contrast-enhanced computed

tomography (diceCT): An emerging tool for rapid, high-resolution, 3-D imaging of

metazoan soft tissues. Journal of Anatomy 228:889–909.

Grillo, O. N., and S. A. K. Azevedo. 2011. Pelvic and hind limb musculature of Staurikosaurus

pricei (Dinosauria: Saurischia). Anais Da Academia Brasileira de Ciencias 83:73–98.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

Sciences 306:431–524.

Hutchinson, J. R. 2001a. The evolution of pelvic osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:123–168.

Hutchinson, J. R. 2001b. The evolution of femoral osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:169–197.

Hutchinson, J. R. 2002. The evolution of hindlimb tendons and muscles on the line to crown-

group birds. Comparative Biochemistry and Physiology - A Molecular and Integrative

Physiology 133:1051–1086.

Hutchinson, J. R., and S. M. Gatesy. 2000. Adductors, abductors, and the evolution of archosaur

locomotion. Paleobiology 26:734–751.

Inger, R. F. 1957. Ecological Aspects of the Origins of the Tetrapods. Evolution 11:373.

78

Jones, S. J., and A. Boyde. 1974. The organization and gross mineralization patterns of the

collagen fibres in sharpey fibre bone. Cell and Tissue Research 148:83–96.

Klinkhamer, A. J., H. Mallison, S. F. Poropat, G. H. K. Sinapius, and S. Wroe. 2018. Three-

Dimensional Musculoskeletal Modeling of the Sauropodomorph Hind Limb: The Effect of

Postural Change on Muscle Leverage. Anatomical Record 301:2145–2163.

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, A. T., and M. R. Gibling. 1996. Stratigraphy and tectonic history of the Upper Devonian

to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology; Volume

32, Number 1 (1996)DO - 10.4138/2076 .

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2017. Reconstructing pectoral

appendicular muscle anatomy in fossil fish and tetrapods over the fins-to-limbs transition.

Biological Reviews. 93:1077-1107

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2018. Evolution of Hindlimb Muscle

Anatomy Across the Tetrapod Water-to-Land Transition, Including Comparisons with

Forelimb Anatomy. Anatomical Record. doi: 10.1002/ar.23997.

Neill, W. T. 1963. Notes on the Alabama Waterdog , Necturus alabamensis Viosca.

Herpetologica 19:166–174.

79

Petermann, H., and M. Sander. 2013. Histological evidence for muscle insertion in extant

amniote femora: Implications for muscle reconstruction in fossils. Journal of Anatomy

222:419–436.

Pierce, S. E., J. A. Clack, and J. R. Hutchinson. 2012. Three-dimensional limb joint mobility in

the early tetrapod Ichthyostega. Nature 486:523–526.

Russell, A. P., and A. M. Bauer. 2008. The appendicular locomotor apparatus of Sphenodon and

normal-limbed squamates. Biology of the Reptilia 21:1–465.

Sanchez, S., V. Dupret, P. Tafforeau, K. M. Trinajstic, B. Ryll, P. J. Gouttenoire, L. Wretman, L.

Zylberberg, F. Peyrin, and P. E. Ahlberg. 2013. 3D Microstructural Architecture of Muscle

Attachments in Extant and Fossil Vertebrates Revealed by Synchrotron Microtomography.

PLoS ONE 8:e56992.

Sanchez, S., D. Germain, A. De Ricqlès, A. Abourachid, F. Goussard, and P. Tafforeau. 2010.

Limb-bone histology of temnospondyls: Implications for understanding the diversification

of palaeoecologies and patterns of locomotion of Permo-Triassic tetrapods. Journal of

Evolutionary Biology 23:2076–2090.

Schoch, R. R. 2014. Amphibian Evolution: The Life of Early Land Vertebrates (M. J. Benton

(ed.)). Wiley Blackwell, Chichester, West Sussex, pp.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Sparreboom, M., and H. Ota. 1995. Notes on the life-history and reproductive behaviour of

Cynops ensicauda popei (Amphibia: Salamandridae). Herpetological Journal 5:310–315.

80

Tanck, E., J. Homminga, G. H. Van Lenthe, and R. Huiskes. 2001. Increase in bone volume

fraction precedes architectural adaptation in growing bone. Bone 28:650–654.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

Walthall, J. C., and M. A. Ashley-Ross. 2006. Postcranial myology of the California newt,

Taricha torosa. 46–57.

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Waters, C. N., R. A. Waters, W. J. Barclay, and J. R. Davies. 2009. A Lithostratigraphical

Framework for the Carboniferous Successions of Northern Great Britain (Onshore). 174 pp.

Witmer, L. M. 1995. The Extant Phylogenetic Bracket and the importance of reconstructing soft

tissues in fossils; pp. 19–33 in Jeff Thomason (ed.), Functional morphology in vertebrate

paleontology. Cambridge University Press.

81

Chapter 4: Midshaft profiles Introduction

One of the greatest difficulties encountered by fin-to-limb researchers is the lack of complete articulated tetrapod specimens in the earliest Carboniferous, and an abundance of disarticulated, isolated elements (Clack, 2006; Mansky and Lucas, 2013). Only recently have fin- to-limb researchers realized that the phenomenon known as “Romer’s Gap” (Romer, 1956;

Coates and Clack, 1995) – a lack of tetrapod fossil material from the late Devonian to the early

Visean – is the result of collection bias (Smithson et al., 2012; Anderson et al., 2015; Clack et al.,

2017, 2018; Smithson and Clack, 2018). With increased efforts at sampling the earliest

Tournaisian (establishment of the TW:eed project in Scotland and Blue Beach Fossil Museum in

Canada) a variety of tetrapod elements are being recovered from Romer’s Gap (Anderson et al.,

2015; Clack et al., 2019; Otoo et al., 2019). The Horton Bluff Formation at Blue Beach Nova

Scotia, which is Tournaisian in age and temporally near the base of Romer’s Gap, is the locality from which the following material was collected.

Blue Beach preserves a well-known footprint fauna, which is suggestive of terrestrial adapted animals living there at the time (Mossman and Grantham, 2008; Mansky and Lucas,

2013). The tetrapod fossils found there comprise an array of isolated elements that appear to represent a range of animals previously known from disparate locations and times, with different suggested locomotor behaviours (Holmes, 1984; Smithson, 1985; Lebedev and Coates, 1995;

Warren and Turner, 2004; Clack and Finney, 2005). Many of these elements appear to be consistent with late Devonian grade digited stem tetrapods such as Acanthostega, Ichthyostega, and Tulerpeton, which are associated with aquatic lifestyles. There are also elements comparable to those of later Visean tetrapods such as the whatcheeriid Pederpes, which was proposed as a

82

more amphibious animal based on symmetry of its digit elements (Clack and Finney, 2005).

Embolomerous-like elements are also found at the site, of which the type material and closely associated genera are believed to be terrestrial based on skull features including reduced or absent lateral lines, large teeth, and “reptilomorph” skulls (Panchen, 1977; Holmes, 1984;

Smithson, 1985). Additionally, Crassigyrinus-like elements from the site could represent animals transitioning from aquatic to terrestrial or from terrestrial back to aquatic lifestyles as the species

Crassigyrinus scoticus from the Visean of Scotland may be a secondarily aquatic animal

(Carroll, 1992b; Herbst and Hutchinson, 2019). Intriguingly, some of the elements have similarities to more than one known taxon, ranging from the Devonian to Carboniferous. For example, YPM.PU.20103 is comparable to both aquatic Tulerpeton and more terrestrial whatcheeriids. The elements preserved at Blue Beach are robust and generally well-preserved in three dimensions and, although they may not be identifiable to the species level, they remain potentially informative about the locomotory behaviour of the animals to which they belonged.

To infer the potential of these animals to live on land, I will conduct a midshaft analysis of the Blue Beach femora. The midshaft (mid-diaphysis) as defined by Laurin (2004) is the narrowest region of the long bone, where the cortical layer is thickest in cross section. The framework for such analyses already exists for the long bones of early tetrapods (Hua and De

Buffrenil, 1996; Laurin et al., 2004, 2011; Germain and Laurin, 2005; Kriloff et al., 2008;

Canoville and Laurin, 2009; Quemeneur et al., 2013). Long bone ossification begins at the midshaft. A bone collar forms around the cartilaginous precursor, and the bone grows longitudinally by endochondral ossification, and in width by the apposition of new bone on the perichondral and periosteal surfaces (Hall, 2007; Padian and Lamm, 2013). The midshaft holds the longest record of a bone’s history because bone tissue deposition begins with a collar around

83

the midshaft. Once maturity is reached the bone does not stop growing. Within the cortical bone osteoblasts in the external periosteal layers deposit bone around the outer circumference of the bone, and internal endosteal layers are remodelled as the bone grows. Cortical bone is the dense outer layer of bone in a long bone, whereas trabecular bone is the porous bone located within the medullary cavity. Both cortical and trabecular bone respond to physiological and mechanical stresses: in cortical bone, responses result in a thickening or thinning of the bone (Lanyon et al.,

1982; Frost, 1990), whereas in trabecular bone there is remodelling of trabecular alignment to better withstand mechanical forces transmitted through the bone (Carter et al., 1989; Biewener et al., 1996; Cowin, 2001; Kivell, 2016). The midshaft of a long bone (i.e., femur, tibia, fibula, humerus, radius and ulna) is reflective of life history because it is a region of bone where deposition happens the longest and the least remodelling occurs (Hall, 2005; Padian and Lamm,

2013). Examining proportions and location of cortical and trabecular bone of a long bone at the midshaft is important for contributing data independent of and in addition to other methods such as external osteological correlates to infer lifestyle. Additionally, the analysis of internal features is important for the Blue Beach material as they may be better preserved than external features.

External osteological correlates of soft tissue are exposed to wear both at the time of deposition and in the present day as the elements are eroded away by tidal action.

The proportion of cortical bone to medullary space, thickness of cortical bone, and transition of cortical to trabecular bone to medullary cavity at the midshaft of the femora may be used as proxies to infer locomotory behaviour of early tetrapods (Currey and Alexander, 1985;

Laurin et al., 2004, 2011; Germain and Laurin, 2005; Kriloff et al., 2008; Canoville and Laurin,

2009; Sanchez et al., 2010; Quemeneur et al., 2013). Cetaceans, for example, have very thin cortical bone and extensive densely packed trabecular bone, so much so that they appear to lack

84

a medullary cavity (Wall, 1983; Laurin et al., 2004). This appears to be the case for many aquatic animals (Wall, 1983; Laurin et al., 2011; Nakajima and Endo, 2013). However, it must be considered that apart from fishes, all other modern aquatic vertebrates are secondarily aquatic, making the aquatic proxies used to compare modern tetrapods with stem tetrapods analogous

(homoplastic), rather than homologous or plesiomorphic.

Previous research using midshaft data compared early fossil tetrapod long bones with those of extant vertebrates in order to assess locomotor behaviour (Wall, 1983; Hua and De

Buffrenil, 1996; Laurin et al., 2004, 2011; Kriloff et al., 2008; Canoville and Laurin, 2009;

Sanchez et al., 2010; Meunier and Laurin, 2012; Quemeneur et al., 2013). Modern taxa are used as references to build a model capable of predicting aquatic or amphibious/terrestrial lifestyles.

The most accurate model (used for the present study) was unable to differentiate between amphibious and terrestrial, but both were easily differentiated from aquatic (Quemeneur et al.,

2013). In previous studies done by Laurin and colleagues, once the reference model was developed fossil taxa were sampled at the midshaft (often by destructive means) for comparison to the extant reference model (Laurin et al. 2004; Germain & Laurin 2005; Kriloff et al. 2008;

Canoville & Laurin 2009; Quemeneur et al. 2013). Girondot and Laurin (2003) developed a program – Bone Profiler- which uses a single slice of the midshaft of a long bone to analyse and determine a compactness profile. The variables used in the predictive model may include some beyond the four main variables Bone Profiler calculates. Size, for example, must be considered in addition to compactness data, as body size variation of animals often influences the size and proportions of the midshaft, and by extension midshaft profile variables. (Laurin et al., 2004;

Germain and Laurin, 2005; Kriloff et al., 2008; Canoville and Laurin, 2009; Quemeneur et al.,

2013).

85

The model for predicting locomotor behaviour based on the femoral midshaft has several limitations but provides a good starting point and comparative results for subsequent analyses.

Limitations include the observations that not all taxa easily fit into this predictive framework

(Laurin et al., 2011). Turtles are different from all the other amniotes sampled and so are excluded from previous works or considered within a separate model (Germain and Laurin,

2005; Kriloff et al., 2008; Quemeneur et al., 2013). The binary nature of the model may be an over simplification of the data but for the purpose of this midshaft analysis the dichotomy is acceptable as it can provide a basic inference of fully aquatic or not fully aquatic behaviours of animals across the water-to-land transition.

I predict that the midshaft profiles of the fossil femora will fall into both categories of aquatic and terrestrial/amphibious, reflective of the diverse array of Devonian-like and

Carboniferous-like tetrapods represented in the Blue Beach fauna (Anderson et al., 2015). I predict that the more Carboniferous-like forms such as the whatcheeriids and embolomeres will display compactness profiles of amphibious to terrestrial animals, whereas the more Devonian- like animals such as Acanthostega and Tulerpeton will have compactness profiles that reflect an aquatic lifestyle.

Abbreviations

Bone Profiler abbreviations:

Global analysis: Minimum asymptote (Min), maximum asymptote (Max), 1/slope (S), transition point (P).

Angular analysis: Minimum asymptote (RMin), maximum asymptote (RMax), 1/slope (RS), transition point (RP).

86

Institutional abbreviations:

Blue Beach Fossil Museum, Hantsport, Nova Scotia, Canada (BBFM); Museum of

Vertebrate Zoology, University of California, Berkeley, USA (MVZ); Nova Scotia Museum of

Natural History, Halifax, Canada (NSM); Redpath Museum, McGill University, Montreal,

Canada (RM); Yale Peabody Museum of Natural History, New Haven, Connecticut, USA,

(YPM).

Materials and Methods

Bone Profiler (Girondot and Laurin, 2003) is a program that, when given an image of a roughly circular, or section of a circular, bone, will analyse 51 concentric zones at 6 degree increments (60 segments per concentric zone) to determine proportions of bone pixels to background or matrix pixels in these areas (Figure 4.1 A). Global variables are calculated by averaging the values of the 60 sectors for each concentric ring. Angular variables are calculated by taking the mean of the concentric rings, but excluding sectors that are outliers. Outliers were defined as “profiles for which at least one of the main four [compactness] parameters was more than four standard deviations away from the mean value” (Laurin et al. 2004). Bone Profiler can automatically assign bone centre, medullary centre, and medullary edge, or they can be assigned manually. All midshafts in these analyses had bone centre, medullary centre and medullary edge assigned automatically. Once these features are assigned the program determines compactness at the midshaft, compactness at the periphery, and overall compactness. It plots the compactness against the distance from the centre and fits a curve graphically representing the minimum asymptote, maximum asymptote, inverse of the slope and transition point of the curve (Figure

4.1 B). Further information regarding the program can be found in Girondot and Laurin (2003).

87

Figure 4.1 Bone Profiler analysis. A, Schematic of the 51 ring and 6-degree increments analysed by bone profiler. B, Graphical representation of the four compactness parameters. Modified from Quemeneur et al. (2013). The angular analysis was used in the femora formula as the previous authors found this to be more accurate than the global analysis (Quemeneur et al., 2013).

The four angular parameters are; Minimum asymptote, which represents the compactness at the least compact part of the bone (generally at the bone centre); maximum asymptote, representing the region of greatest compactness of the bone (generally the cortical bone); the transition point is the position within the midshaft at which the most abrupt change in bone compactness occurs; and 1/slope is the inverse of the slope on the graph at the transition point and usually reflects the width of the transition zone (Girondot and Laurin, 2003). For the Binary model (aquatic or not aquatic) which is used in my analysis, the angular value of the minimum asymptote (Rmin) is the most informative variable about lifestyle (Quemeneur et al., 2013).

Animals with higher compactness near the bone centre, or more trabecular bone throughout the medullary cavity are associated with more aquatic lifestyles.

88

Table 4.1 Femora used for midshaft profiles, and their historical locomotor behaviour associations. Specimen Element Comparable To Previous Locomotor Behaviour Association

NSM.005.GF.045.048-A Femur Embolomere Amphibious to Terrestrial

RM.20.6711 Femur Acanthostegid Aquatic

NSM.007.GF.004.630 Femur Tulerpetontid (whatcheeriid) Aquatic

NSM.005.GF.045.044 Distal Whatcheeriid (Ossinodus-like) Amphibious to Femur Terrestrial

NSM.005.GF.045.047 Femur Whatcheeriid (Ossinodus-like) Amphibious to Terrestrial

YPM.PU.20103 Femur Whatcheeriid (revised) Aquatic to Amphibious

The femora used in the present study are NSM.005.GF.045.044 (distal femur),

NSM.005.GF.045.047 (the distal and proximal ends), NSM.005.GF.045.048-A (whole femur),

NSM.007.GF.004.630 (whole femur), RM.20.6711 (whole femur), YPM.PU.20103 (whole femur). The Family identification and associated lifestyles are shown in Table 1. Some femora were excluded (NSM.005.GF.045.042, YPM.PU.23550) because their internal structures had either significant crushing, were of very similar density to the infill matrix, or had an abundance of high-density inclusions. Portions of incomplete femora where a midshaft could be identified were used (NSM.005.GF.045.047, NSM.005.GF.045.044). Midshaft profiles of a platypus femur

(MVZ:32885) and a cat femur were used to ground truth the accuracy of the Bone Profiler program, the binary formula provided by Quemeneur et al. (2013; available in the supplementary material associated with that text), and methods used to threshold the bones. They were not the only animals of terrestrial and aquatic lifestyles used, as the extensive dataset in Quemeneur et al. (2013) included many aquatic, amphibious, and terrestrial taxa to create the model formula used here. The platypus and cat femora were used to visually represent compactness profiles of

89

aquatic and terrestrial taxa in this study but additional variables from other animals are in

Quemeneur et al. (2013).

The femora were scanned in an Xradia Versa 510 (Carl Zeiss, Germany) at the University of Calgary with settings of 140kV, 72µA, no filter, and magnification 0.4x. The scans produced stacks of 1928 - 2412 images approximately 1000 pixels wide by 1024 pixels tall, and voxel sizes ranging between 21.9-32.2 µm.

Midshafts were identified by loading the image stack into Data Viewer (Bruker), and examining it for the narrowest circumference. The midshaft region identified was anywhere between 20-200 transaxial images, and once identified the images were further examined in the transaxial plane to determine which image in that 20-200 image stack had the thickest cortical bone. Once selected, the single midshaft image was thresholded using ImageJ (Bethesda,

Maryland, USA). Before thresholding, the contrast of the image was manipulated by eye so as to remove the “air” pixels and enhance the distinction between bone and matrix. Following this, the autothreshold function was applied with “ignore black” and “show threshold values in log window” options selected. The output from this final operation included a sample of results from all of the automatically thresholded binary images. The image that appeared to most accurately reflect the unaltered image of the midshaft in terms of maintaining the cortical thickness and trabecular structure at the bone centre was selected for further analysis. The thresholding function and value (i.e. Huang 122 (Huang and Wang, 1995)) was recorded and then applied to the midshaft (Table 4.2). The midshaft image was then de-noised by applying a median filter of kernel size ranging

90

Table 4.2. ImageJ threshold functions and values of midshaft slices

Specimen Threshold function Threshold Value

NSM.005.GF.045.044 Huang (Huang and Wang, 122 1995)

NSM.005.GF.045.047-D Shanbhag (Shanbhag, 1994) 32191

NSM.007.GF.004.630 Shanbhag (Shanbhag, 1994) 190

NSM.005.GF.045.048-A Huang (Huang and Wang, 135 1995)

RM.20.6711 RenyiEntropy (Kapur et al., 27179 1985)

between 1 and 2. This resulted in a binary image of the midshaft that was then exported to Bone

Profiler (Figure 4.2 A-C).

One specimen, YPM.PU.20103, was segmented using Dragonfly 4.0 (ORS Montréal) because ImageJ was unable to autothreshold the image and beam hardening was more easily corrected for using Dragonfly. The fileset was loaded into Dragonfly, non bone containing image slices were removed from the stack, and the image was cropped to reduce the dataset size. The entire dataset was segmented and thresholded. First a median filter of kernel size 3 (the smallest kernel size for the median filter in Dragonfly) was applied, followed by a mean shift of kernel size 2 (and X and Y maximal distance of 5). Then an arithmetic function was applied to convert the images to 32bit float so that the following processes could be applied: a polynomial filter (of the third degree), followed by a Distance of Gaussian (DoG) filter (first sigma 0.5, second sigma

4.0). Once the filters were applied to the dataset it was thresholded (defining a specific range of pixel intensities representing bone to the exclusion of matrix and air) to isolate a region of interest (ROI) of primarily the trabecular bone (see Appendix D for further details). The cortical

91

bone of the original dataset was thresholded by hand as the filtered dataset could not effectively automatically threshold the cortical bone and trabecular bone (see Appendix D steps 10-12). The cortical and trabecular bone had to be thresholded separately because beam hardening artifacts made the matrix near the cortical bone the same density as trabeculae in the middle. Once each region was thresholded the cortical and trabecular ROIs were united (Boolean operations) to create a binary image stack of the isolated bone structure. After this, the slice selected as the midshaft by methods outlined above (looking for the narrowest region along the shaft, where the cortical bone was thickest in transaxial views) was exported as a binary file to Bone Profiler.

Bone Profiler automatically analysed the exported images and determined the section centre, medullary centre, and ontogenetic centre (Figure 4.2 D). The global and angular outputs of the scaled logistic model were drawn and fitted (Figure 4.2 E); the global and angular results of all femora are included in Appendix B.

Based on the formula of Quemeneur et al. (2013; Supplementary Table S3) the angular parameter RMin was most informative for differentiating between the aquatic, or amphibious/terrestrial lifestyles of amniote femora. As RMin is related to trabecular structure near the centre of the bone, aquatic animals would be likely to have higher RMin values (i.e., more trabecular bone in the medullary space) than terrestrial animals. The parameter RMin obtained from each femur in this analysis was plugged into the formula of Quemeneur et al. (2013) to determine the lifestyle of the animals associated with the fossils.

Quemeneur et al. (2013) examined the midshafts of 155 animal femora including reptile, mammal, and bird, from which they created a binary model that accurately predicted lifestyle

(aquatic or amphibious/terrestrial) in all the extant femora tested. The Binary model relied on the

92

Figure 4.2. Example of how a midshaft (RM.20.6711) was analysed in bone profiler. A, original midshaft image. B, midshaft after contrast was adjusted. C, binary midshaft after threshold applied. D, binary image uploaded into Bone Profiler and bone parameters automatically generated. E, global and angular output from Bone Profiler.

Compactness parameter Rmin. When plugged into the binary formula, RMin resulted in either an output of 0 or 1, 0 coding aquatic, and 1 coding amphibious/terrestrial lifestyle.

Results

The four main compactness parameters determined by the Bone Profiler software

(Girondot and Laurin, 2003) are given as numerical values in Table 4.3 as well as plotted on a graph with the axes of compactness vs distance from bone centre (Figure 4.3).

93

Figure 4.3. Bone Profiler graphs illustrating the compactness profiles of the Blue Beach fossil Femora, with a platypus (MVZ:32885) included for comparison representing an aquatic bone profile, and cat representing terrestrial.

NSM.005.GF.045.044

The midshaft slice was determined to be 1156 and the Huang (Huang and Wang, 1995) threshold (ImageJ) was applied at a value of 122 (Figure 4.4 A, B). RMin was 0.1538543 (SD

0.2019223) and when run through the binary formula of Quemeneur et al. (2013) received a score of 0, inferring an aquatic lifestyle.

NSM.005.GF.045.047-Distal

The midshaft slice used was 647, and a Shanbhag (Shanbhag, 1994) threshold of 32191 was applied (Figure 4.4 C, D). RMin was 0.3806298 (SD 0.1854969) which resulted in a score of

0 when inserted into the binary formula.

94

NSM.007.GF.004.630

Midshaft slice was 969, and a Huang (Huang and Wang, 1995) threshold of 135 was applied (Figure 4.4 E, F). RMin was 0.2926539 (SD 0.2323235) and received a score of 0 the binary formula was applied.

Table 4.3. Angular Compactness Profiles of Blue Beach Femora Midshafts

NSM.005.GF.045.044

Mean SD

Rmin 0.1538453 0.2019223

Rmax 0.9932303 0.0175256

RS 0.0950134 0.0547641

RP 0.5537147 0.138318

NSM.005.GF.045.047-D

Rmin 0.3806298 0.1854969

Rmax 0.9908298 0.0342943

RS 0.044946 0.049484

RP 0.6928861 0.1469062

NSM.007.GF.004.630

Rmin 0.3039398 0.2416092

Rmax 0.960207 0.0824888

RS 0.1420611 0.1558052

RP 0.0105079 0.2574734

NSM.005.GF.045.048-A

Rmin 0.2461127 0.1623108

Rmax 0.9793683 0.0557052

RS 0.0714518 0.108942

RP 0.6996397 0.1126213

95

RM.20.6711

Rmin 0.4127885 0.146311

Rmax 0.9542483 0.1082802

RS 0.0600025 0.1026328

RP 0.7773617 0.232480

YPM.PU.20103

RMin 0.5391226 0.0905507

RMax 0.9945045 0.0348535

RS 0.0249126 0.0320448

RP 0.8530897 0.0685409

NSM.005.GF.045.048-A

Midshaft slice was 849, and the Shanbhag (Shanbhag, 1994) threshold was applied at a value of 27419 (Figure 4.4 G, H). RMin was calculated to be 0.2461127 (SD 0.1623108) and scored a 0 in the binary model, again inferring an aquatic lifestyle.

RM.20.6711

Slice 1306 was identified as the midshaft, and the threshold Renyientropy (Kapur et al.,

1985) was applied at a value of 27179 (Figure 4.4 I, J). RMin was 0.4127885 (SD 0.146311), which scored a 0 in the binary model.

YPM.PU.20103

Slice 1050 of the cropped dataset was selected as the midshaft (Figure 4.4 K, L), established from the three-dimensional view of Dragonfly as the narrowest region of the bone.

The true midshaft (where cortical bone was thickest in transaxial view) was not used because of a crack. Instead, the image used was the closest slice to the images with the thickest cortical bone

96

Figure 4.4. Unmodified µCT (A,C,E,G, I, and K) and thresholded binary images (B,D,F,H, J, and L) of the midshafts of fossil femora. unaffected by the crack. Details of its segmentation and thresholding are described above. RMin was 0.5391226 (SD 0.0905507) and was assigned 0, as aquatic.

Graphical results from bone profiler show that all the Blue Beach fossil femora have a high compactness near the bone centre, more similar to that of the aquatic platypus than the terrestrial cat. All compactness graphs of the femora, except NSM.007.GF.004.630, have a distinct S-shaped curve which illustrates a more rapid change in compactness at approximately the point of transition between trabecular bone and cortical bone than in NSM.007.GF.004.630, which has a much more gradual transition.

Additional angular outputs of each six-degree increment are included in Appendix B, and the four main parameters of both global and angular compactness are included in Appendix C.

97

Discussion

These results find that the diverse tetrapod fauna of Blue Beach, which includes

Devonian-like and Carboniferous-like forms, are aquatic. This suggests that later Carboniferous forms, such as other whatcheeriids that are proposed to be some of the early vertebrates with more terrestrial capabilities (such as Ossinodus, Pederpes, and embolomerous animals; Holmes,

1984; Smithson, 1985; Clack and Finney, 2005) may not have been as terrestrial as previously thought. Importantly, this also supports the conclusion that the fin-to-limb transition was not contemporaneous to the water to land transition. Animals appear to be developing digited limbs for use in an aquatic environment rather than for an increased tendency to move across land. This study cannot determine how the animal was moving in its aquatic environment, but confirms that they were not terrestrial, and the evolution of digited limbs was not prompted by the movements of fish-like finned animals onto land.

The compactness graphs illustrate a range of compactness profiles where all the curves of the femora appear distinct from one another. Apart from NSM.007.GF.004.630, all the femora have clear S-shaped angular (blue) analysis curves (Figure 4.3), but there is much variation in minimum compactness, location of transition point with respect to the bone centre, and slope of the curve at the transition point. Animals that are similar morphotypes, such as

NSM.005.GF.045.044 and NSM.005.GF.045.047, have visibly different compactness at the bone centre, and curves reflecting a difference between medullary and cortical bone transition.

NSM.005.GF.045.047 has a rapid, relatively discrete, transition between medullary and cortical bone, whereas NSM.005.GF.045.044 has a gradual transition. Such differences could be related to age rather than locomotor behaviour because trabecular bone volume changes through ontogeny (Tanck et al., 2001) and may be reflected in variation of trabecular thickness and, by

98

extension, compactness of trabecular bone at the midshaft. Intriguingly, the femur

NSM.007.GF.004.630, the only embolomerous femur in this sample, is the most different from all of the other femora, including the cat and platypus. This animal has the least clear differentiation between cortical bone and trabecular bone (gradual rather than abrupt transition), which could partially be an effect of weathering. Alternatively, it could represent an older animal that has undergone extensive remodelling. Histological sectioning of the sample could reveal further detail, and support or reject this idea. Some variation of the compactness curves may represent insignificant variations within individuals, but there may also be signals from age, use of limbs (paddling, kicking, gliding), benthic vs pelagic, or primary vs secondarily aquatic lifestyles. Without the graphs (only the numerical values were published) of the 155-animal dataset of Quemeneur et al. (2013) I have no comparative basis for inferring what these differences mean.

A platypus femur underwent the same analyses as these fossil femora and coded a 0, or aquatic. Platypods do walk in a terrestrial manner and could be considered amphibious.

Quemeneur et al. (2013) defines an aquatic tetrapod as having advanced specialized morphology specific to its aquatic lifestyle, and where all foraging takes place in an aquatic environment. So, while a platypus may traverse terrestrial territory, it is still classified as aquatic. This could mean that the Blue Beach tetrapods were aquatic to a similar degree as a platypus, and they may have been capable of short terrestrial excursions. A cat femur was also scanned and run through the

Bone Profiler program, and the RMin produced a value of 1 when plugged in to the binary formula of Quemeneur et al (2013), inferring a terrestrial lifestyle, as expected. Whereas the

Blue Beach femora are most similar to the platypus femur, they remain very different from both platypus and cat, suggesting there are differences between these animals not captured in this

99

analysis. The Blue Beach femora may be signalling that early tetrapod fossil limbs do not fit into the model provided by Quemeneur et al. (2013) and are not comparable to modern limb bones.

Inferences from internal morphology of the Blue Beach femora could be made more robust by the inclusion of three dimensional data taking the whole bone into account. Whereas the midshaft is where bone has been deposited for the longest time, the trabecular alignment of the proximal neck and head in mammals such as humans shows where forces associated with the impact of terrestrial locomotor patterns act across the bone. Trabeculae align in a pattern similar to the struts and beams of a crane in order to withstand the forces placed on them (for a review see Cowin 2001; Kivell 2016). Further examination of trabecular structure across these early tetrapod long bones could be useful to determine where forces were acting on the limb bones and yield more insight into gait and locomotor behaviour of the animals.

Conclusion

What is clear from the femoral elements at Blue Beach is that there was a variety of different animals locomoting in an aquatic environment during the early Tournaisian. I would expect a combination of terrestrial and aquatic animals to mirror the diversity in external morphology of Devonian-like and Carboniferous-like forms; however, all of the femora I analysed using Bone Profiler (Girondot and Laurin, 2003) are inferred to belong to aquatic animals. Perhaps the diverse group of animals moved through their aquatic environment in different ways, but the present study is silent on this question. These results support the hypothesis that the development of limbs was not preceded by the movement of fish-like vertebrates bearing limbs into more terrestrial environments. Limbs appear to be relatively well developed before they are used for locomotion on land.

100

A compactness profile analysis is a good place to start when attempting to determine the locomotory behaviours of fossil tetrapods from Blue Beach. The inclusion of more three- dimensional data, which is used in more modern fossils (such as hominin fossils; Ryan and

Ketcham, 2002a, 2002b, 2005; Zeininger et al., 2016; Su and Carlson, 2017; Dunmore et al.,

2019), but not rigorously examined until recently in non mammalian tetrapods (Bishop et al.,

2018; Plasse et al., 2019) may provide further details about habitial loading regimes on the limbs of Blue Beach tetrapods. Use of three-dimensional data in the study of internal structure of the limb bones of early tetrapods could result in greater disparity between categories of locomotory behaviours and yield a more robust result.

Literature Cited

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Biewener, A. A., N. L. Fazzalari, D. D. Konieczynski, and R. V Baudinette. 1996. Adaptive

changes in trabecular architecture in relation to functional strain patterns and disuse. Bone

19:1–8.

Bishop, P. J., S. A. Hocknull, C. J. Clemente, J. R. Hutchinson, A. A. Farke, B. R. Beck, R. S.

Barrett, and D. G. Lloyd. 2018. Cancellous bone and theropod dinosaur locomotion. Part

I—an examination of cancellous bone architecture in the hindlimb bones of theropods.

PeerJ 6:e5778.

Canoville, A., and M. Laurin. 2009. Microanatomical diversity of the humerus and lifestyle in

lissamphibians. Acta Zoologica 90:110–122.

Carroll, R. L. 1992. The Primary Radiation of Terrestrial Vertebrates. Annual Review of Earth

101

and Planetary Sciences 20:45–84.

Carter, D. R., T. E. Orr, and D. P. Fyhrie. 1989. Relationships between loading history and

femoral cancellous bone architecture. Journal of Biomechanics 22:231–44.

Clack, J. A. 2006. The emergence of early tetrapods. Palaeogeography, Palaeoclimatology,

Palaeoecology 232:167–189.

Clack, J. A., C. E. Bennett, D. K. Carpenter, S. J. Davies, N. C. Fraser, T. I. Kearsey, J. E. A.

Marshall, D. Millward, B. K. A. Otoo, E. J. Reeves, A. J. Ross, M. Ruta, K. Z. Smithson, T.

R. Smithson, and S. A. Walsh. 2017. Phylogenetic and environmental context of a

Tournaisian tetrapod fauna. Nature Ecology & Evolution 1.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Clack, J. A., L. B. Porro, and C. E. Bennett. 2018. A Crassigyrinus -like jaw from the

Tournaisian (Early Mississippian) of Scotland. Earth and Environmental Science

Transactions of the Royal Society of Edinburgh 108:37–46.

Clack, J. A., C. E. Bennett, S. J. Davies, A. C. Scott, J. E. Sherwin, and T. R. Smithson. 2019. A

Tournaisian (earliest Carboniferous) conglomerate-preserved non-marine faunal assemblage

and its environmental and sedimentological context. PeerJ 2019:e5972.

Coates, M. I., and J. A. Clack. 1995. Romer’s gap: tetrapod origins and terrestriality. Bulletin Du

Muséum National d’Histoire Naturelle 4:373–388.

Cowin, S. C. 2001. The false premise in Wolff’s law; pp. 30-1-30–16 in Bone Mechanics

Handbook, Second Edition.

102

Currey, J. D., and R. M. N. Alexander. 1985. The thickness of the walls of tubular bones. Journal

of Zoology 206:453–468.

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

Dunmore, C. J., T. L. Kivell, A. Bardo, and M. M. Skinner. 2019. Metacarpal trabecular bone

varies with distinct hand-positions used in hominid locomotion. Journal of Anatomy

235:45–66.

Frost, H. M. 1990. Skeletal structural adaptations to mechanical usage (SATMU): 1. Redefining

Wolff’s Law: The bone modeling problem. The Anatomical Record 226:403–413.

Germain, D., and M. Laurin. 2005. Microanatomy of the radius and lifestyle in amniotes

(Vertebrata, Tetrapoda). Zoologica Scripta 34:335–350.

Girondot, M., and M. Laurin. 2003. Bone Profiler: A tool to quantify, model, and statistically

compare bone-section compactness profiles. Journal of Vertebrate Paleontology 23:458–

461.

Hall, B. K. 2005. Bones and Cartilage: Developmental and Evolutionary Skeletal Biology.

Elsevier, San Diago, 1–760 pp.

Herbst, E. C., and J. R. Hutchinson. 2019. New insights into the morphology of the

Carboniferous tetrapod Crassigyrinus scoticus from computed tomography. Earth and

Environmental Science Transactions of the Royal Society of Edinburgh 109:157–175.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

103

Sciences 306:431–524.

Hua, S., and V. De Buffrenil. 1996. Bone histology as a clue in the interpretation of functional

adaptations in the thalattosuchia (reptilia, crocodylia). Journal of Vertebrate Paleontology

16:703–717.

Huang, L.-K., and M.-J. Wang. 1995. Image thresholding by minimizing the measures of

fuzziness. Pattern Recognition 28:41–51.

Kapur, J. N., P. K. Sahoo, and A. K. C. Wong. 1985. A new method for gray-level picture

thresholding using the entropy of the histogram. Computer Vision, Graphics, & Image

Processing 29:273–285.

Kivell, T. L. 2016. A review of trabecular bone functional adaptation: What have we learned

from trabecular analyses in extant hominoids and what can we apply to fossils? Journal of

Anatomy 228:569–594.

Kriloff, A., D. Germain, A. Canoville, P. Vincent, M. Sache, and M. Laurin. 2008. Evolution of

bone microanatomy of the tetrapod tibia and its use in palaeobiological inference. Journal of

Evolutionary Biology 21:807–826.

Lanyon, L. E., A. E. Goodship, C. J. Pye, and J. H. MacFie. 1982. Mechanically adaptive bone

remodelling. Journal of Biomechanics 15:141–154.

Laurin, M., M. Girondot, and M.-M. Loth. 2004. The evolution of long bone microstructure and

lifestyle in lissamphibians. Paleobiology 30:589–613.

Laurin, M., A. Canoville, and D. Germain. 2011. Bone microanatomy and lifestyle: A

descriptive approach. Comptes Rendus - Palevol 10:381–402.

104

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, T. A., and M. R. Gibling. 1996. Stratigraphy and tectonic history of the Upper Devonian

to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology 32:13–38.

Meunier, F. J., and M. Laurin. 2012. A microanatomical and histological study of the fin long

bones of the Devonian sarcopterygian Eusthenopteron foordi. Acta Zoologica 93:88–97.

Nakajima, Y., and H. Endo. 2013. Comparative Humeral Microanatomy of Terrestrial,

Semiaquatic, and Aquatic Carnivorans using Micro-Focus CT Scan. Mammal Study 38:1–8.

Otoo, B. K. A., J. A. Clack, T. R. Smithson, C. E. Bennett, T. I. Kearsey, and M. I. Coates. 2019.

A fish and tetrapod fauna from Romer’s Gap preserved in Scottish Tournaisian floodplain

deposits. Palaeontology 62:225–253.

Padian, K., and E.-T. Lamm. (Eds.). 2013. Bone Histology of Fossil Tetrapods: Advancing

Methods, Analysis, and Interpretation. University of California Press, Berkeley, California,

298 pp.

Panchen, A. L. 1977. On Anthracosaurus russelli Huxley (Amphibia: Labyrinthodontia) and the

Family Anthracosauridae. Philosophical Transactions of the Royal Society B: Biological

Sciences 279:447–512.

Plasse, M., E. Amson, J. Bardin, Q. Grimal, and D. Germain. 2019. Trabecular architecture in

105

the humeral metaphyses of non‐avian reptiles (Crocodylia, Squamata and Testudines):

Lifestyle, allometry and phylogeny. Journal of Morphology jmor.20996.

Quemeneur, S., V. de Buffrénil, and M. Laurin. 2013. Microanatomy of the amniote femur and

inference of lifestyle in limbed vertebrates. Biological Journal of the Linnean Society

109:644–655.

Romer, A. S. 1956. The Early Evolution of Land Vertebrates. Proceedings of the American

Philosophical Society 100:157–167.

Russell, A. P., and A. M. Bauer. 2008. The appendicular locomotor apparatus of Sphenodon and

normal-limbed squamates. Biology of the Reptilia 21:1–465.

Ryan, T. M., and R. A. Ketcham. 2002a. Femoral head trabecular bone structure in two

omomyid primates. Journal of Human Evolution 43:241–263.

Ryan, T. M., and R. A. Ketcham. 2002b. The three-dimensional structure of trabecular bone in

the femoral head of strepsirrhine primates. Journal of Human Evolution 43:1–26.

Ryan, T. M., and R. A. Ketcham. 2005. Angular orientation of trabecular bone in the femoral

head and its relationship to hip joint loads in leaping primates. Journal of Morphology

265:249–263.

Sanchez, S., D. Germain, A. De Ricqlès, A. Abourachid, F. Goussard, and P. Tafforeau. 2010.

Limb-bone histology of temnospondyls: Implications for understanding the diversification

of palaeoecologies and patterns of locomotion of Permo-Triassic tetrapods. Journal of

Evolutionary Biology 23:2076–2090.

Shanbhag, A. G. 1994. Utilization of Information Measure as a Means of Image Thresholding.

106

CVGIP: Graphical Models and Image Processing 56:414–419.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Smithson, T. R., and J. A. Clack. 2018. A new tetrapod from Romer’s Gap reveals an early

adaptation for walking. Earth and Environmental Science Transactions of the Royal Society

of Edinburgh 108:89–97.

Smithson, T. R., S. P. Wood, J. E. A. Marshall, and J. A. Clack. 2012. Earliest Carboniferous

tetrapod and arthropod faunas from Scotland populate Romer’s Gap. Proceedings of the

National Academy of Sciences of the United States of America 109:4532–7.

Su, A., and K. J. Carlson. 2017. Comparative analysis of trabecular bone structure and

orientation in South African hominin tali. Journal of Human Evolution 106:1–18.

Tanck, E., J. Homminga, G. H. Van Lenthe, and R. Huiskes. 2001. Increase in bone volume

fraction precedes architectural adaptation in growing bone. Bone 28:650–654.

Tibert, N. E., and D. B. Scott. 1999. Ostracodes and Agglutinated Foraminifera as Indicators of

Paleoenvironmental Change in an Early Carboniferous Brackish Bay, Atlantic Canada.

PALAIOS 14:246.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

Wall, W. P. 1983. The correlation between high limb-bone density and aquatic habits in Recent

mammals (cetaceans, pinnipeds). Journal of Paleontology 57:197–207.

107

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Zeininger, A., B. A. Patel, B. Zipfel, and K. J. Carlson. 2016. Trabecular architecture in the StW

352 fossil hominin calcaneus. Journal of Human Evolution 97:145–158.

108

Chapter 5: Trabecular Analysis Introduction

Locomotion in early tetrapods was determined historically using proxies such as limb bone proportions, degree of ossification of the limb elements, or digit number (Coates and Clack,

1990; Clack, 2001; Laurin et al., 2004; Warren and Turner, 2004; Hall, 2007; Warren, 2007;

Coates et al., 2008; Sallan et al., 2010; Sanchez et al., 2010; Padian and Lamm, 2013). None of these proxies has been experimentally or quantitatively tested, so it is unclear if they truly reflect biomechanics or physiology associated with locomotor capability of the animal. An increasing number of biomechanical studies on early tetrapods (Sanchez et al., 2010; Pierce et al., 2012;

Diogo et al., 2016; Molnar et al., 2017, 2018; Nyakatura et al., 2019) inferred locomotor behaviour and the evolution of musculature associated with terrestrial locomotion. One study, by

Pierce et al. (2012), estimated the joint movement potential of Ichthyostega, but their estimates of cartilage and soft tissues rely on comparisons to extant taxa with “similar” bodies.

Unfortunately, for early tetrapods there are no similar extant taxa, which makes such comparisons significant extrapolations. Another method used to estimate locomotor capabilities of early tetrapods is soft tissue (specifically muscle) reconstruction (Bryant and Russell, 1992;

Witmer, 1995; Hutchinson and Gatesy, 2000; Hutchinson, 2001a, 2001b, 2002; Carrano and

Hutchinson, 2002; Diogo et al., 2016; Molnar et al., 2017, 2018), but this too has many shortfalls as described in Chapter Three. Other recent investigations examined internal midshaft anatomy, which has an created an abundance of two-dimensional data in the literature (Germain and

Laurin, 2005; Ward et al., 2006; Kriloff et al., 2008; Canoville and Laurin, 2009; Sanchez et al.,

2010; Laurin et al., 2011; Meunier and Laurin, 2012; Quemeneur et al., 2013; Chapter Four of this work). Recently, a few studies have examined three-dimensional volumes of the internal

109

structure of limb bones of early tetrapods (Sanchez et al., 2012, 2013, 2016), comparing virtual methods of histology to traditional histology (Sanchez et al., 2012), providing information about potential soft tissue interactions with implications for location and orientation of musculature

(Sanchez et al., 2013) and patterns of development and ossification (Sanchez et al., 2016).

Trabecular bone, also known as cancellous or spongey bone, is recognized as a useful material for the study of behaviour in fossil animals. Behaviourally moderated repetitive forces acting on bone induce bone remodelling (Nakajima and Endo, 2013; Kivell, 2016; Su and

Carlson, 2017; Bishop et al., 2018; Plasse et al., 2019). The organization of trabecular remodelling provides additional and potentially more sensitive information about forces acting on the bone. While fleshy (muscle to bone) attachments do not always leave external osteological correlates (Bryant and Seymour, 1990; Petermann and Sander, 2013) the force they exert on the bone may leave internal osteological correlates in the form of trabecular orientation

(see Figure 5.1).

In the fields of medicine and paleoanthropology there is an abundance of literature on the quantification of mechanical loading effects on internal bone structure of in extant and extinct mammalian material (Biewener et al., 1996; Ulrich et al., 1999; Ketcham and Ryan, 2004; Ryan and Ketcham, 2005; Carlson et al., 2008; Wallace et al., 2013; Kivell, 2016; Reznikov et al.,

2016; Williams-Hatala et al., 2018; Kivell et al., 2018; Tsegai et al., 2018; Georgiou et al.,

2019), but these methods have only recently been used in non-mammalian fossils. As of 2019 only two studies have examined such data: one on dinosaurs (Bishop et al., 2018) and another on non-avian reptiles (Plasse et al., 2019).

110

Figure 5.1. Coronal slice of the proximal femur of a juvenile cat. Highlighted in red is highly aligned trabecular structure that is indicative of the force transmitted from pelvis to femur in response to walking in a terrestrial environment (i.e., under gravitational forces). Two kinds of cells are critical to the formation and remodelling of bone: osteoblasts, which deposit bone, and osteoclasts, which resorb bone (Eurell and van Sickle, 1998).

Inconsistent patterns of loading during growth are known to stimulate osteogenesis (Hall, 2005), and promote trabecular realignment to counteract dynamic forces associated with repetitive loading and unloading (von Meyer 1867, Roux 1881, Wolff 1892; for a review see Kivell 2016).

Figure 5.1 illustrates where the loading “beams” (trabeculae) arch from the femoral head of a cat femur, to a thickened region of cortical bone below the femoral neck (highlighted in red). They are aligned to best distribute forces acting across the joint.

111

Wolff’s law

Wolff’s Law, now more accurately termed Bone Functional Adaptation (BFA), was made famous by Wolff (1892), but was originally conceived by Roux (1881), from observations by von Meyer (1867) (Cowin, 2001; Kivell, 2016). BFA uses the general theory that trabecular bone remodels to counteract repetitive forces acting on it (Hall 2005), but not the mathematical formulas described by Wolff in 1892, which are now recognized as inaccurate (Cowin, 2001;

Kivell, 2016) because there is not a one-to-one relationship between stress trajectories and trabecular orientation (Cowin, 2001). Analyses of the organization and alignment of trabecular bone (anisotropy), in conjunction with bone volume fraction, are often used to make inferences about BFA, and thus, behaviours involving repetitive force in extinct animals (Biewener et al.,

1996; Fajardo and Müller, 2001; Ryan and Ketcham, 2005; Wallace et al., 2013; Kivell et al.,

2018; Williams-Hatala et al., 2018; Dunmore et al., 2019; Plasse et al., 2019).

Anisotropy

Isotropy describes the homogenous alignment of elements making up a three-dimensional structure. Anisotropy describes the non-homogenous alignment of components making up a three-dimensional structure. In trabecular structure, anisotropy refers to the non-random organization of a three-dimensional network of trabeculae. Studies examining trabecular structure in fossils have primarily focused on mammals (Goldstein et al., 1991; Smit et al., 1997;

Ryan and Ketcham, 2002b, 2002a; Carlson et al., 2008; Wallace et al., 2013; Nakajima and

Endo, 2013; Zeininger et al., 2016; Kivell, 2016; Su and Carlson, 2017; Tsegai et al., 2018;

Georgiou et al., 2019). Kivell (2016) effectively summarizes much of the early research on trabecular bone analysis and its potential uses for fossil specimens: a) trabecular bone remodels throughout life in a more predictable way than external bone features; b) trabecular bone (in

112

conjunction with cortical bone) reflects actual behaviours (locomotor behaviours in this case), rather than potential behaviours; and c) there does not appear to be an allometric relationship between size of the animal and degree of anisotropy.

Equivocal results, however, confound the study of trabecular bone and its capability to describe behaviour. Carlson et al. (2008) observed that mice that moved through linear, turning, and control environments did not have differing degrees of anisotropy in the distal femoral metaphysis. Ryan and Ketcham (2005) observed weak differences in anisotropy of the trabecular bone in the femoral head of leaping vs non leaping primates. Finally, Pontzer et al. (2005) observed that trabecular orientation of the distal femur did reflect changes in load orientation across the knee joint in guinea fowl. It is suggested (Ryan and Ketcham, 2002a; Su and Carlson,

2017) that results pertaining to the degree of anisotropy and trabecular orientation are extremely sensitive to the location of the volume of interest (VOI) within the bone. Therefore, trabecular analytical methods need to be further refined and standardized, so that appropriate regions are examined for variables that discriminate between load bearing behaviours.

Anatomists have described trabecular orientation for over a century (von Meyer, Roux,

Wolff); however, paleontologists have only recently had access to methods for non-destructive examination of the internal structure of fossil bone in three dimensions (microcomputed tomography and synchrotron). Previously, researchers had to cut, grind, and destroy large sections of material to visualize the structures inside fossils. Therefore, destructive sampling was used sparingly, and the internal structures of fossil material remained relatively unknown. With the advent, decreasing size, and cost of microcomputed tomography (microCT) scanners a new generation of paleontologists can visualize and analyze internal structures in three dimensions.

Use of CT scans on extant animals and in the field of medicine means methods for segmenting

113

cortical bone from trabecular bone (and medullary cavity) are well documented and computer programs have been developed to automate this process (Buie et al., 2007; Kohler et al., 2007).

Fin-to-limb

The transition from aquatic to terrestrial environments leads to multiple changes in the forces acting on limb bones. For example, fish fins are exposed to forces (buoyancy, water pressure, gravity) acting in all directions underwater, whereas terrestrial limbs are exposed to gravitational and ground reaction forces acting in a more vertical plane. When the animal locomotes across a terrestrial environment forces act inconsistently on the bones. This inconsistency of forces (walking) causes trabecular remodelling (Hall, 2005). Pectoral fins are oriented lateral to the trunk rather than ventral to the trunk as terrestrial limbs interacting with substrate are. Pelvic fins are ventral projections that are oriented similarly to a pelvic limb however, most fish do not actually locomote with their fins, rather their tail and axial body undulations provide propulsive power. Therefore, the forces exerted on proximal fin elements by musculature should pull in different locations and directions than forces exerted by musculature on proximal elements of terrestrial limbs and this should be reflected in the location, direction and magnitude of anisotropy in the limb bone. Aquatic limbs should have muscle forces acting equally in all directions to move the limb against hydrodynamic forces, whereas terrestrial muscles working to move a limb against the force of gravity should exhibit forces of greater magnitude in the dorsoventral plane. Terrestrial limbs with directional gravitational and ground reaction forces acting on them should have highly anisotropic regions at the proximal and distal ends, oriented towards the point of articulation with the element transmitting gravitational or ground reaction forces. Aquatic limbs with musculature inducing trabecular realignment should have a lesser magnitude of high anisotropy in locations where muscle interacts with bone.

114

With this background, I predict that Blue Beach limb bones belonging to more terrestrially adapted animals will have greater anisotropy within the trabecular structure along loading trajectories. Specifically, I expect regions of high anisotropy in terrestrial limbs to be located where gravitational forces and ground reaction forces act across joints (i.e., from pelvis to proximal femur, and across the knee). I also expect more trabecular alignment associated with the topological location of muscle attachment in terrestrial animals, because the muscles must exert enough force to move the mass of the limb in combination with the force of gravity (i.e., weight). For more aquatic limb bones I expect less anisotropy because of less directional joint loading and muscle forces counteracting gravity because I expect the fewer directional forces applied to aquatic limb bones to be ineffective at promoting trabecular remodelling.

Abbreviations

Blue Beach Fossil Museum, Hantsport, Nova Scotia, Canada (BBFM); Musée d’Histoire

Naturelle de Miguasha, Québec, Canada (MHNM); Museum of Vertebrate Zoology, University of California, Berkeley, USA (MVZ); Nova Scotia Museum of Natural History, Halifax, Canada

(NSM); Redpath Museum, McGill University, Montreal, Canada (RM); University of Michigan

Museum of Zoology, Ann Arbor, Michigan, USA (UMMZ); Yale Peabody Museum of Natural

History, New Haven, Connecticut, USA (YPM); Museum für Naturkunde Berlin, Germany

(ZMB).

Ani.M, Mode of frequency of anisotropy. BV/TV, bone volume fraction, Ct.Ar/Tt.Ar, cortical area fraction.

115

Locality

At the time of deposition in the Carboniferous, Blue Beach was a high energy aquatic environment occasionally exposed to marine conditions (Tibert and Scott, 1999; Mansky and

Lucas, 2013). Material was likely transported from elsewhere and disarticulated prior to deposition. Anderson et al. (2015) noted a diverse range of Devonian-like and Carboniferous-like morphotypes present at the site, suggesting a diversity of locomotor behaviours. The animals include Acanthostega-like, Tulerpeton-like, whatcheerid, Crassigyrinus-like, Proterogyrinus- like, and Eoherpeton-like. While Acanthostega is definitively aquatic, the animals such as the whatcheeriid Pederpes, as well as Proterogyrinus and Eoherpeton are thought to be more amphibious (Holmes, 1984; Smithson, 1985; Clack and Finney, 2005). The site is dated as mid to early-late Tournaisian in age, approximately coeval with the CM palynozone in Britain

(Utting et al., 1989; Martel et al., 1993; Martel and Gibling, 1996; Waters et al., 2009), which is currently under review and may be older (TW:eed project; Bennett et al., 2017; Clack et al.,

2017; Otoo et al., 2019).

Present day Blue Beach is located between Hantsport and Avonsport, Nova Scotia,

Canada, near the Bay of Fundy, a region known for extreme (greater than 10m) tides. The fossils are eroded away daily. Surface details are well preserved despite wear from the high energy environment at both the times of deposition and collection. Critical to this analysis, much of the material has excellent internal preservation.

Materials and Methods

Fossil material used in this study included two complete and three partial femora, two tibiae, and two fibulae, in addition to one humerus, one radius, and one ulna. These elements

116

represent a range of Devonian to Carboniferous grade animals (see Chapter Two for descriptions). The elements underwent an internal analysis of trabecular structure to assess locomotor behaviour. Extant limb bones studied here include the leopard gecko Eublepharis, spiny tailed lizard Uromastyx, marine iguana Amblyrhynchus cristatus (FMNH UF41558), hellbender salamander Cryptobranchus alleganiensis (FMNH UF10881), platypus

Ornithorhynchus anatinus (MVZ MVZ32885), femur of a (juvenile) housecat Felis, humerus of an otter Lutra lutra (ZMB MAM59069), and a femur of the turtle Elseya dentata (UMMZ herps

203554). I also sampled a humerus of the extinct tetrapodomorph fish Eusthenopteron foordi

(MHNM 06-2674A). These elements were obtained to provide a range of material with varying locomotor behaviours for comparison with Blue Beach fossils. Eublepharis, Uromastyx, and

Felis samples were obtained from specimens used for previous research or teaching at the

University of Calgary. The platypus, marine iguana, otter, turtle, and hellbender salamander were obtained from Morphosource (https://www.morphosource.org/). The specimen of

Eusthenopteron was borrowed from the Natural History Museum at Miguasha, Quebec.

New tetrapod material from Blue Beach was collected by Chris Mansky of the BBFM, some of which was prepared or partially prepared by Sonja Wood. Previously collected material was borrowed for scanning from the collections BBFM, NSM, RM, and YPM. The material was borrowed by JSA and KL and further prepared by KL if needed using Paleotools Microjacks 1-3

(Brigham, Utah) and pin vice, and reinforced with carbowax and polyvinyl acetate (PVA).

Acetone was used to remove previously applied glue as well as to help differentiate the fossil from matrix to facilitate preparation. Once fully removed from the matrix the material was scanned in an Xradia Versa 510 (Carl Zeiss, Germany) by KL at the University of Calgary with settings of 140kV, 72µA, no filter, and magnification 0.4x. The scans resulted in image stacks

117

approximately 1001 pixels wide by 1024 high (except for any wide-angle scans), ranging from

20 – 35µm voxel size. Cylindrical or elongate elements were scanned and vertically stitched in the Xradia Scout and Scan program and consisted of 2-3 stitched scans. Two humeri from YPM were too wide to fit in the Xradia scanner and so were scanned using the SkyScan 1173 (Bruker,

Belgium) at the University of Calgary. The bones were preferentially scanned in the Xradia µCT

X-ray scanner, as the air/no filter better differentiated fossil material from matrix, and ring artefacts were negligible.

Binary files of the scans were saved and converted to TIFF image stacks, which were converted to image sequences in ImageJ (NIH, Maryland). The resulting image sequence was exported to Dragonfly (ORS Montréal) for cortical and trabecular segmentation and analysis (the following methods are graphically documented in Appendix D). The image stack was smoothed using the median filter of kernel size 3-5 and a mean shift filter of 5-7 (mean shift was always kernel size of 2 higher than the median filter). A region of interest (ROI) encompassing the background (air and packing material) of the scan was thresholded and overwritten with the density of the bone, which was determined using Dragonfly’s “probe” feature. Density of fossil bone is not reflective of the density of the bone in life, therefore this number was simply used as a reference to overwrite the background region of interest. An Arithmetic filter was then applied to the image stack to convert the 16-bit TIFFS to 32-bit float files, which was required for the later Distance of Gaussian (DoG) filter application. Following the Arithmetic filter, a Polynomial function of the third degree was applied. Then a Distance of Gaussian filter was applied with the first sigma ranging between 0.5 and 1.0 and the second sigma ranging from 2.0 to 4.0. The filtered dataset was then thresholded so that the range of values most accurately representing the trabecular bone were isolated and selected as an ROI. Filtering made the images less noisy, and

118

reduced the beam hardening artefacts, but in some instances this filtering process could not accurately capture the thick cortical bone of the image stack. When this happened the cortical bone of the image stack was thresholded separately using the output from the median-mean-shift filtered dataset. Following this, the median-mean-shift cortical bone ROI and median-mean-shift- arithmetic-polynomial-DoG trabecular ROI were united (Boolean operations) to create an accurate representation the whole bone structure. Once an ROI which accurately represented the trabecular and cortical structures was created, the background ROI was subtracted from it.

The merged ROI underwent the process of cortical and trabecular bone segmentation in the bone analysis module of Dragonfly. The Kohler method (Kohler et al., 2007) was used because there was more manual control over the differentiation of cortical and trabecular areas, which helped for the unusually shaped specimens.

Once the cortical and trabecular bone were segmented the program output 4 ROIs: cortical area, trabecular area, cortical bone and trabecular bone. The trabecular bone ROI was primarily used for the remainder of the analysis. First, the trabecular bone ROI underwent a multiROI analysis (6-connected) to remove any small areas of high pixel intensity remaining that were not connected to trabecular or cortical bone. This was done by isolating the largest single volume from the multi-ROI analysis, creating the trabecular network. The largest volume, representing just the trabecular bone, was exported to a new ROI (named “trab used” or “real trab”).

After segmenting and isolating the trabecular bone, the cortical area and trabecular area were united (Boolean operations) to form an “infill” ROI – an ROI representing the entire volume including interior spaces of the bone. This was required to run a segmentation analysis to

119

calculate the average anisotropy, trabecular thickness, cortical thickness, cortical area fraction

(Ct.Ar/Tt.Ar), bone volume (BV), and total volume (TV) measurements. Note that as this was the first release of the Bone Analysis module (Dragonfly 4.0) BV and TV were reversed so bone volume fraction (BV/TV) was calculated in Microsoft Excel. The average trabecular widths were used as a basis for the radius of influence and spacing used in the following trabecular analysis.

This study is unlike others that analyse trabecular orientation in bones because it measures anisotropy at a much smaller scale over the entire bone. Most fossil analyses (Ryan and

Ketcham, 2002b, 2002a; Zeininger et al., 2016; Su and Carlson, 2017; Bishop et al., 2018c;

Plasse et al., 2019) have used several specifically located volumes of interest (VOIs) where they measure one average anisotropy measurement for each specific volume (for example the femoral neck Ryan & Ketcham 2002b; Ryan & Ketcham 2002a). In this study smaller VOIs from the trabecular network of the entire bone were analysed to pick up fine details of trabecular bone orientation. Volumes of interest (created by spacing and sampling parameters) needed to be large enough that the trabecular bone surface was not mapped as an area of orientation (i.e., VOIs had to be larger than the width of a trabecula). The spacing and sampling parameters (VOIs) for creating vector and scalar maps of anisotropy were therefore at least three times the average trabecular width. Elements with average trabecular widths of 167µm or less were analysed with sampling and spacing of 500µm, all other elements had sampling and spacing of 900µm. The marine iguana had remarkably thick trabeculae, so the sampling and spacing parameters were

2000µm. Of the fossil limb bones, only the humerus YPM.PU.23545 and distal femur

NSM.005.GF.045.047 had average trabecular widths greater than 167µm.

Removal of internal fossae in the whatcheeriid humerus was done slice by slice to ensure the filters did not incorporate that area as part of the bone. Due to the unusual shape of the

120

humerus, the bone analysis module in Dragonfly had difficulty separating cortical from trabecular bone. Therefore, to ensure the trabecular network did not include any of the cortical bone thicker regions of cortical bone along the ectepicondyle and cortical bone lining fossae were segmented out slice by slice.

Multiple boxes (area box) encompassing the trabecular areas of interest (whole bone, or slices of bone) were created to define where to sample VOIs and map the trabecular network. All quantitative analyses were run on the whole bone area box of interest. After the box of interest was created the mapping tab from the bone analysis sub menu was selected. The surface normal algorithm (Dragonfly, ORS) was selected and “trabROI” or “real trab” chosen as the ROI to perform the analysis on within the area box of interest (Appendix D steps 18-19). Spacing for all fossil specimens was 500µm because average trabecular width for all other fossils had trabeculae narrower than 167µm. The two mentioned above with thicker trabeculae, YPM.PU.23545 and

NSM.005.GF.045.047, has spacing of 900µm. Radius of influence was set at the same value

(overlap of 0.5) as the spacing for each specimen. No previous studies have determined ideal methods for the analysis of so many small VOIs to my knowledge. If the spacing was of smaller or equal size to the width of a single trabecula there would be an abnormally high degree of anisotropy because it would capture the surface of a single trabecula. Conversely, if the spacing was too large the anisotropy would be reflective of a larger average area and therefore have error due to averaging effects and a weaker signal. Anisotropy evaluation was calculated using surface normal (Dragonfly ORS, Montréal), the mapping was projection based, and a smoothing of 2 was used; these parameters were suggested to me when I was trained by the Dragonfly (ORS

Montréal) developers.

121

The maps computed from these parameters included a scalar anisotropy heat map (a two- dimensional image stack with slices matching the spacing steps), and a three-dimensional vector map showing the direction of trabecular regions within the specified volumes across the bone.

The colour of the vector arrows indicates the degree of anisotropy. Arrow heads were removed to simplify the vector maps presented here (see results).

The radius of influence often extended beyond where the trabeculae and cortical bone were segmented from each other. This leads to large oriented surfaces at the edge of the trabecular bone and creates bias towards high anisotropy. This results in two-dimensional scalar maps of anisotropy with areas of high anisotropy extending beyond the edge of the trabecular bone (Figure 5.2 B). Highly anisotropic values are erroneously mapped around areas where thickened trabecular bone attaches to the cortex (Figure 5.2 A, B). While I do not know of a way to correct for this effect in the three-dimensional vector mapping, it can be easily corrected in the two-dimensional scalar anisotropy map (Figure 5.2 C, D). First, the trabecular area of the segmentation is inverted (Figure 5.2 C). The two-dimensional scalar map is converted to nearest neighbour form before the inverted trabecular background is used to overwrite all values outside of the true trabecular area with a value of 0 (thus removing most of the sample areas composed of the cortical-trabecular “cut”). The scalar anisotropy map is then converted back to a cubic interpolation (Figure 5.2 D), and the histogram of frequencies of anisotropy analysed from this created an altered, more accurate map. While effects from trabecular-cortical edges in scalar maps can be corrected for, anisotropy vector maps should be viewed with caution, especially when most regions of anisotropy are located near the edges of trabecular bone. Cross referencing between the two-dimensional heat map of anisotropy can help to conclude if areas of high

122

Figure 5.2. Transaxial slice of fossil tibia NSM.014.GF.036.004. trabecular structure is highlighted in pink. A shows that trabecular structure is left with aligned edges where it is cut from the trabecular structure in segmentation. B shows how this effects anisotropy calculations at the edges where these cuts get analysed and produce areas of high anisotropy. C shows the inverted trabecular area (grey-blue) that will be overwritten on the anisotropy heat map. D shows the more accurate overwritten anisotropy heat map. anisotropy near the trabecular-cortical “cut” are a factor of segmentation, or a true biological Figure 5.3. Anisotropy vector maps and histograms of extant terrestrial humeri. A-D Eublepharis humerus. B, vector map of all anisotropy. C, Vector map123 of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis humerus. F-I, Uromastyx humerus. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx humerusFigure 5.4.

feature of the bone (see Appendix E for detailed description of this process; steps 8-14).

This process was repeated for all fossil bones. Extant animal bones generally only had the median and mean shift filters applied before going on to the bone analysis and cortical and trabecular segmentation. If the median and mean shift filters did not yield an accurate ROI, the procedures outlined above were used. After the median-mean shift filters were applied the resulting ROI was segmented in the bone analysis program, and the cortical and trabecular bone were automatically segmented. The trabecular bone ROI was then analysed in the same manner as described above from the creation of the “infill” ROI (see Appendix E for detailed extant segmenting procedure).

A correlation Principal Components Analysis (PCA) was run in PAST (Hammer, Harper and Ryan, 2001) on the raw output variables Proximodistal length of the bone (PD), bone volume fraction (BV/TV), average cortical area fraction (Ct.Ar/Tt.Ar), and mode of the frequency of anisotropy (Ani.M; mode is used because the data will not be

Table 5.1. Variables of whole bone analysis

Specimen BV/TV CT.Ar/T Cortical Trabecular PD Ani.M t.Ar Thickness Thickness (cm) (µm) (µm)

NSM.007.GF.004.630 0.440986 0.34 265.04 152.1 6.6 0.0650

NSM.005.GF.045.061 0.464487 0.52 348.2 127.9 3.8 0.1412

NSM.005.GF.045.112 0.351262 0.58 362.44 110.57 4.0 0.1833

NSM.005.GF.045.047 - 0.516936 0.58 533.67 158.17 3.8 0.2500 distal

NSM.005.GF.045.048-A 0.458478 0.37 387.16 132.47 6.1 0.0500

NSM.007.GF.004.733 0.344645 0.46 241.24 102.37 3.4 0.0167

NSM.014.GF.036.002 0.453296 0.5 450.09 154.75 3.3 0.0500

124

NSM.014.GF.036.004 0.413546 0.48 471.65 132.9 3.0 0.0333

RM.20.5222 0.411031 0.45 262.58 105.06 3.4 0.1000

NSM.005.GF.045.044 0.455864 0.43 592.35 179.17 5.0 0.0167

NSM.005.GF.045.047 - 0.285265 0.18 166.33 99.82 2.5 0.0643 proximal

YPM.PU.23545 0.374153 0.17 237.27 175.06 5.2 0.0235

Terrestrial Femora

Felis 0.495491 0.63 667.27 189.93 9.6 0.5429

Uromastyx 0.398976 0.8 260.96 118.67 1.9 0.6830

Eublepharis 0.542776 0.92 296.49 103.85 1.9 0.4923

Amphibious Femora

Amblyrhynchus 0.654598 0.84 1468.32 501.79 7.8 0.3333

Ornithorhynchus 0.510064 0.69 535.19 272.88 3.2 0.3286

Terrestrial Humeri

Uromastyx 0.412424 0.85 222.87 79.37 1.5 0.6769

Eublepharis 0.654809 unknown unknown 110.07 1.6 0.6154

Amphibious Humeri

Ornithorhynchus 0.722861 0.81 797.53 unknown 3.5 0.4538

Amblyrhynchus 0.634388 0.77 1445.25 488.81 8.0 0.2000

Lutra 0.580385 0.72 1290.44 223.57 8.5 0.4286

Aquatic Femora

Elseya 0.60018 0.73 unknown unknown 3.1 0.0200

Aquatic Humeri

Eusthenopteron 0.347286 0.21 239.54 166.18 3.7 0.1571

exportable as a CSV file until the next update in January). A correlation rather than variance- covariance PCA was used because the variables were of different measurements and magnitudes.

125

The correlation PCA standardized the variables before analysing them to avoid distortion of the data. Additionally, Dragonfly could not calculate an average value for elements with more variable trabecular and cortical thicknesses (Ornithorhynchus humerus, Elseya femur,

Eublepharis humerus) so average trabecular and cortical thickness were not used in the PCA.

Results

A note on the frequency curves: 0.0 is by far the highest frequency of anisotropy in all the analysed bones, and although not visualized (except in Figure 5.5 O) it makes up the highest frequency in the histogram. The high frequency of an anisotropy of 0 is a result of overwriting the scalar anisotropy map’s inverted trabecular area with 0. The team at Dragonfly are working to rectify the problem. Anisotropy is measured on a scale of 0.0 to 1.0, where 0.0 represents perfectly isotropic organization and 1.0 represents anisotropic organization within the volume analysed.

The average trabecular thickness, average cortical thickness, bone volume fraction, proximodistal length (PD), and cortical area fraction are reported in Table 5.1.

Extant terrestrial taxa

Humeri

Eublepharis (Figure 5.3 A-D) has scattered areas of high orientation. These areas of orientation are primarily at the proximal and distal ands, and they appear to have little pattern to the alignment. The vectors at the proximal and distal ends have no clear alignment and are heterogenous in their directions. There are a few anisotropic vectors near the proximal midshaft, the result of an isolated trabecula extending across the shaft. There is no clear trend in the

126

Figure 5.5. Anisotropy vector maps and histograms of extant terrestrial humeri. A-D Eublepharis humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis humerus. F-I, Uromastyx humerus. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx humerus. frequencies of anisotropy in Eublepharis (Figure 5.3 E), there are almost no regions of anisotropy greater than 0.8, and from 0.0 to 0.8 the frequencies are relatively constant.

Uromastyx (Figure 5.3 F-I) shows many highly anisotropic and aligned vectors at proximal and distal ends of the bone. The proximal humerus has a concentration of highly anisotropic vectors in a fan shape near the articulation point with the shoulder girdle. The distal humerus consist of highly aligned and oriented vectors across the entire broad surface of the distal bone, again in a fan shape. Uromastyx has a distinct frequency curve peaking between 0.6 and 0.8 (Figure 5.3 J). There is a far clearer and higher peak in the histogram of frequency of anisotropic regions in the humerus of Uromastyx than any other extant bone examined.

127

Femora

Eublepharis (Figure 5.4 A-D) has several areas of high anisotropy along the shaft, neck, and head. This distal end of the bone shows some orientation primarily in a vertical direction.

The proximal end of the bone has regions of orientation that strongly align with the neck and head. The frequency of anisotropy in Eublepharis (Figure 5.4 E) appears to increase slightly at higher degrees of anisotropy.

Uromastyx (Figure 5.4 F-I) has almost no regions of high orientation along the femoral shaft. There is a large concentration of aligned highly anisotropic regions at the distal end. The distal vectors are aligned in a fan shape compared to the vertical orientation of Eublepharis. The proximal end of the bone has fewer areas of high anisotropy than the distal end, but several of them are clearly oriented in a way that suggests an effect from the interaction between the proximal femur and pelvis. There is a slightly higher frequency of anisotropy in Uromastyx compared to Eublepharis. In Uromastyx it ranges 0.6-0.8 (Figure 5.4 J). The frequency of anisotropy here has a curve with a clear peak as opposed to Eublepharis.

The proximal femoral head and neck of Felis (Figure 5.4 K-N) have the highest concentration of areas with high anisotropy in the bone. This is consistent with what is observed in Figure 5.1 and as expected due to its terrestrial locomotor behaviour. The distal end of the bone has fewer areas with high anisotropy. The vectors in the femoral neck are consistent with where forces transmitted through the pelvis are transmitted through the femoral head to a

128

Figure 5.6. Anisotropy vector maps and histograms of extant terretrial femora. A-D Eublepharis femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Eublepharis femur. F-I, Uromastyx femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Uromastyx femur. K-N, Felis femur. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Felis femur. thickened area of cortical bone distal to the neck. The peak frequencies of anisotropy appear between 0.5 and 0.7 in Felis (Figure 5.4 O).

129

Extant aquatic taxa

Humeri

Cryptobranchus (Figure 5.5 A-E) has a few regions of high anisotropy concentrated at the edge of the proximal articulating surface. Alignment of the high anisotropy vectors is not consistent with what is expected of a limb reacting to joint loading under gravity. Nor does it display any consistency in alignment that would suggest force exerted from muscle attachment is influencing the trabecular structure. The most frequent value of anisotropy is between 0.4 and

0.5; however, few regions were examined due to the large trabecular thickness but small bone size (makes a VOI at least three times the average trabecular width almost as wide as the bone).

It is unclear if what was observed in the medullary cavity of Cryptobranchus limb bones was actually trabecular structure. Therefore, the limb bones of Cryptobranchus are not used for further comparative analysis as a representative of expected trabecular alignment in an aquatic animal.

Ornithorhynchus (Figure 5.5 F-I) had highly oriented regions across the whole bone, but they were poorly aligned. A proximodistal slice aligning with the broad distal end (Figure 5.5 I), illustrates regions of higher anisotropy (darker red) near proximal and distal articulation surfaces; however, they appear in random directions with no alignment. The Ornithorhynchus frequency histogram (Figure 5.5 J) shows highest frequencies of anisotropy range from 0.2-0.8. The histogram has relatively consistent frequencies, with no obvious peak.

Amblyrhynchus (Figure 5.5 A-D) has proximal and distal ends with almost no regions of high anisotropy. The midshaft is where high anisotropy vectors are congregated. There is no clear alignment to the high anisotropy vectors present in any section of the bone. A slight proximodistal pattern may be observed at the shaft when the vector map is thresholded to

130

Figure 5.7. Anisotropy vector maps and histograms of extant aquatic humeri. A-D Cryptobranchus humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Cryptobranchus humerus. F-I, Ornithorhynchus humerus. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Ornithorhynchus humerus. K-N, Amblyrhynchus humerus. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Amblyrhynchus humerus. anisotropy of 0.65 and above (Figure 5.5 M). However, a midshaft slice (Figure 5.5 N) reveals internal vectors with less proximodistal alignment were obscured in Figure 5.5 M. The

Amblyrhynchus frequency histogram (Figure 5.5 E) appears opposite that seen in the terrestrial cat. This suggests instances of high anisotropy are more frequent in terrestrial animals and instances of low anisotropy are more frequent among the aquatic animals. The most frequent

131

measurements of anisotropy to peak below 0.5, were around 0.2 in Amblyrhynchus and few instances of anisotropy between 0.8 and 1.0 are observed.

Femora

Vector maps of the femur of Cryptobranchus (FMNH UF10881) (Figure 5.6 A-D) show very few regions of anisotropy, and even fewer with high anisotropy. Anisotropy mappings and calculations for this sample are suspect because the limb was so small and trabeculae so large that the size of the region of interest likely lead to significant trabecular-cortical segmentation edge effects. Cryptobranchus (Figure 5.6 E) has no anisotropic regions of 0.8 or greater, and the most common value of anisotropy is below 0.4. Again, the material within the marrow cavity of the bone may not have been trabecular bone. It was very thick and amorphous. Values of anisotropy that were only 0.0, 0.4, and 0.5, which lead me to suspect assessing trabecular anisotropy in Cryptobranchus is not possible.

Ornithorhynchus (Figure 5.6 F-I) has most of its high regions of orientation at the proximal end. However, a proximodistal slice aligned with the broad axis of the bone reveals that the high orientation is concentrated at the periphery of the trabecular structure. The midshaft area shows a moderate amount of orientation, but with no clear alignment. Distally there are fewer regions of high anisotropy without clear alignment. The frequency of anisotropy histogram in

Ornithorhynchus (Figure 5.6 J) has a slight curve, peaking around 0.4.

Amblyrhynchus (Figure 5.6 K-N) has very few regions of high orientation at the proximal or distal ends of the femur. The proximal midshaft appears to have the greatest concentration of regions of high anisotropy aligned in a proximodorsal direction.This corresponds well with the attachment area of the puboischiofemoralis internus of Iguana (Molnar et al., 2019; figure 2 u).

132

Figure 5.8. Anisotropy vector maps and histograms of extant aquatic femora. A-D Cryptobranchus femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above. D, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. E, Anisotropy frequency histogram of Cryptobranchus femur. F-I, Ornithorhynchus femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above. I, Vector map of the middle region in the coronal plane with anisotropy 0.65 and above. J, Anisotropy frequency histogram of Ornithorhynchus femur. K-N, Amblyrhynchus femur. L, vector map of all anisotropy. M, vector map of anisotropy of 0.65 and above. N, vector map of the mid region of the bone in the coronal plane with anisotropy 0.65 or higher. O, Anisotropy frequency histogram of Amblyrhynchus femur.

Amblyrhynchus (Figure 5.6 O) has a downward trend in the values of frequency of anisotropy. A slight peak is present between 0.2 and 0.3 before values of higher anisotropy decline. There are greater frequencies of anisotropy below 0.4 than above it.

133

Figure 5.9. Anisotropy vector maps and histogram of a Eusthenopteron humerus. B, vector map of all anisotropy. C, vector map of anisotropy 0.65 and above in the flexor half of the bone. D, vector map of anisotropy 0.65 and above in the extensor half of the bone. E, anisotropy frequency histogram of Eusthenopteron humerus.

Fossil bones

Humerus

A humerus (MHNM 06-2674A) of Eusthenopteron (Figure 5.7) was analysed to provide an example of an aquatic fossil specimen. Regions of high anisotropy are located all over the periphery of the trabecular bone and lacking within the internal trabecular area. The regions of high anisotropy display no consistent pattern in their alignment. The anisotropy histogram

(Figure 5.7 D) peaks around 0.2 and declines so that there are more instances of anisotropy below 0.5 than above 0.5.

YPM.PU.23545, a whatcheeriid humerus (Figure 5.8 A-D), has regions of high (0.65 or greater) anisotropy spaced all around the periphery of the trabecular bone. There is an orientation pattern visible near the entepicondylar foramen, but this is likely a function of cortical-trabecular bone segmentation because the foramen is completely lined with finished bone. The anisotropy histogram (Figure 5.8 E) has a steep curve peaking at 0 with very few frequencies of anisotropic values between 0.6 and 1.0.

134

Figure 5.10 Anisotropy vector maps and histograms of fossil forelimb elements. A-D, YPM.PU.23545 whatcheerid humerus. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of YPM.PU.23545. F-I, NSM.007.GF.004.733 Tulerpeton-like ulna. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.007.GF.004.733. K-N, NSM.014.GF.036.002 Acanthostega-like radius. L, vector map of all anisotropy. M, vector map of anisotropy 0.65 and above in the flexor half of the bone. N, vector map of anisotropy 0.65 and above in the extensor half of the bone. O, Anisotropy frequency histogram of NSM.014.GF.036.002.

Ulna

NSM.007.GF.004.733, a Tulerpeton-like ulna (Figure 5.8 F-I), has anisotropy vector maps that display a high number of regions where anisotropy is 0.65 or greater; however, there

135

does not appear to be any pattern to the orientation of these regions. The anisotropy histogram

(Figure 5.8 J) curves down from 0, with very few occurrences of anisotropy values between 0.8-

1.0.

Radius

The acanthostegid radius NSM.014.GF.036.002 (Figure 5.8 K-N) has anisotropy vector maps that show several regions of high anisotropy around the midshaft. The high areas of anisotropy near the extensor surface appear random; however, on the flexor surface the vectors appear to radiate out from the midshaft, near the long furrow on the bone. This furrow may be associated with a flexor muscle between the radius and digits. Much of the rest of the anisotropy occurs at the periphery of the trabecular bone. The anisotropy histogram (Figure 5.8 O) of this element shows a gradual decline from 0 to 0.7, and then very low or no frequency of regions of anisotropy 0.7-1.0.

Femora

NSM.007.GF.004.630 (Figure 5.9 F-I) has regions of high anisotropy concentrated around the edges where trabecular bone meets cortical bone. Oddly, the direction of the highly anisotropic edge orientation is not consistent with the proximodistal orientation where cortical bone and trabecular bone are cut off in segmentation of other fossil elements. Rather, a great amount of the orientation is anterioposteriorly and dorsoventrally oriented (although it is random and lacks pattern), perhaps relating to the increasing thickness of trabecular bone as it nears the cortical bone. The midshaft region of the bone has the highest concentration of highly anisotropic bone, especially near the remains of the adductor blade. However, there is no

136

Figure 5.11. Anisotropy vector maps and histograms of fossil femora. A-D, NSM.005.GF.045.048-A embolomerous femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.048-A. F-I, NSM.007.GF.004.630 indet femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.007.GF.004.630. consistent alignment. NSM.007.GF.004.630 has the most frequent anisotropy values of 0.2 or lower, after which the histogram curves downwards (Figure 5.9 J).

NSM.005.GF.045.048-A (Figure 5.9 A-D) has anisotropy vector maps that reveal no regions of high anisotropy within the internal portion of the bone. The regions of high anisotropy on the vector map are restricted to the periphery of the trabecular structure, and therefore likely an artefact of segmentation edges. The blade appears to have areas of high segmentation all roughly oriented vertically; however, this could also be an artefact of segmentation or the crack along the adductor blade. The frequency histogram (Figure 5.9 E) has a high frequency of low anisotropy values that become less frequent as degree of anisotropy increases.

137

NSM.005.GF.045.044, the distal end of an Ossinodus-like femur (Figure 5.10 A-D) has mapped vectors of anisotropy 0.65 or higher located sparsely and near the bone surface. This is likely an artifact of trabecular-cortical segmentation edges. The midshaft has a concentration of high anisotropy vectors but they lack obvious alignment. The anisotropy histogram (Figure 5.10

E) slopes down from 0 with highest frequencies of anisotropy values 0.1 or lower, and very low frequencies of anisotropic values 0.7 or higher.

NSM.005.GF.045.047, another distal end of a whatcheeriid femur (Figure 5.10 F-I), has a higher number of regions with high anisotropy compared to the above femora; however, most are concentrated near the cortical trabecular interface and therefore are associated with cortical- trabecular segmentation edges. There are a few regions of high anisotropy in internal trabecular structure, midway between the dorsal and ventral surface, approximately in line with the adductor crest. They do not have a clear pattern. The anisotropy histogram (Figure 5.10 J) has a curve that peaks between 0.2 and 0.4, consistent with extant aquatic animals, but different from all the other Blue Beach femora mentioned so far in that it does not slope down from 0. The peak frequency of anisotropy is between 0.2 and 0.4, the highest frequency of anisotropy above 0.2 than all of the other fossil femora yet considered.

The proximal end of the whatcheeriid femur NSM.005.GF.045.047 (Figure 5.10 K-N) has anisotropy vector maps with areas of high anisotropy primarily around the edges of trabecular structure. They are likely cortical-trabecular bone segmentation artefact. Highest frequency of anisotropy is between 0.0 and 0.2, with fewer frequencies of higher values of anisotropy (Figure 5.10 O).

138

Figure 5.12. Anisotropy vector maps and historgrams of partial fossil femora. A-D, NSM.005.GF.045.044 Ossinodus-like distal femur. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.044. F-I, NSM.005.GF.045.047 whatcheerid distal femur. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.005.GF.045.047. K-N, NSM.005.GF.045.047 whatcheerid proximal femur. L, vector map of all anisotropy. M, vector map of anisotropy 0.65 and above in the flexor half of the bone anisotropy 0.65 and above in the flexor half of the bone. N, vector map of anisotropy 0.65 and above in the extensor half of the bone. O, Anisotropy frequency histogram of NSM.005.GF.045.047.

Fibulae

The crassigyrinid fibula NSM.005.GF.045.112 (Figure 5.11 A-D) has high anisotropy vectors around the edges of the trabecular area. No pattern is clearly discernable although some

139

of the yellow vectors (anisotropy nearer 0.64 than 1.0) appear to radiate fan-like from the midshaft, and there is a congregation of higher anisotropy regions near the crack, which is to be expected and not reflective of a true biological phenomenon. The radiating yellow vectors could be indicative of long proximodistally oriented areas of connectivity between trabecular and cortical bone. The anisotropy histogram (Figure 5.11 E) is similar to that of the distal femur

NSM.005.GF.045.047; there is a distinct peak, rather than an immediate downward slope from 0.

The peak appears around 0.2, before the frequency of higher anisotropy values sharply declines.

Very few instances of anisotropy between 0.8 and 1.0 are observed. This is consistent with the pattern seen in extant aquatic humeri and femora.

The Pederpes-like fibula NSM.005.GF.045.061 (Figure 5.11 F-I) has a great number of high anisotropy vectors, and in flexor view some appear to be radiating out from the centre of the element, although this is obscured by random areas of higher anisotropy. More importantly, some of the radiating vectors appear to be internal to the trabecular-cortical segmentation edge therefore may be a true biological feature reflecting a remodelling of trabecular structure to adapt to forces acting on the bone. The anisotropy histogram (Figure 5.11 J) remains moderately flat between 0 and 0.2 before gradually sloping down as decreasing frequencies of high anisotropic bone are observed. No instances of anisotropy values between 0.8 and 1.0 are observed.

140

Figure 5.13. Anisotropy vector maps and histograms of fossil fibulae. A-D, NSM.005.GF.045.112 Crassigyrinus-like fibula. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of NSM.005.GF.045.112. F-I, NSM.005.GF.045.061 Pederpes-like fibula. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.005.GF.045.061.

Tibiae

NSM.014.GF.036.004 (Figure 5.12 F-I), a whatcheeriid tibia excellently preserved in three dimensions, has an accumulation of anterioposteriorly-oriented high anisotropy vectors near midshaft of the anterior edge in flexor view and extending all the pay to the posterior edge in extensor view. Most of these appear to be at the periphery of the trabecular area; few of the regions extend internally. They do appear highly aligned and may be indicative of trabeculae connecting to cortical bone along broad regions of anteroposteriorly directed bone. This is not consistent with expectations of alignment associated with gravitational impact. The anisotropy histogram (Figure 5.12 J) shows a curve down from 0, where there are higher frequencies of low anisotropy and fewer frequencies of high anisotropy.

141

Figure 5.14. Anisotropy vector maps and histograms of fossil tibiae A-D, RM.20.5222 Proterogyrinus- like tibia. B, vector map of all anisotropy. C, Vector map of anisotropy 0.65 and above in the flexor half of the bone. D, Vector map of anisotropy 0.65 and above in the extensor half of the bone. E, Anisotropy frequency histogram of RM.20.5222. F-I, NSM.014.GF.036.004 “type 2” whatcheerid tibia. G, vector map of all anisotropy. H, Vector map of anisotropy 0.65 and above in the flexor half of the bone. I, Vector map of anisotropy 0.65 and above in the extensor half of the bone. J, Anisotropy frequency histogram of NSM.014.GF.036.004.

RM.20.5222 (Figure 5.12 A-D) has an abundance of high anisotropy vectors running along the crack in this bone, so these must be viewed with caution as they are not a true reflection of the orientation of the trabecular bone. Cracks must be viewed with caution because the DoG filter enhances contrast at edges. A crack has significant differences in pixel intensity values where bone/matrix and air/glue meet, so the contrast between them is amplified. This can obscure the more subtle differences between bone and matrix such that matrix can be labelled as bone where it meets the air/glue of a crack. Anisotropy of the extensor half of the bone shows no clear pattern of oriented regions, apart from those associated with the crack. The extensor half of the bone displays slightly more patterned regions of anisotropy primarily radiating out in a proximodistal direction from the midshaft of the bone. However, a slice through the interior of

142

the bone shows that most of these aligned regions are associated with the periphery of the trabecular area. The anisotropy frequency histogram (Figure 5.12 E) slopes down from 0 where the highest frequencies of anisotropy observed are less than 0.2, and few frequencies of anisotropy between 0.8 and 1.0 are observed.

Principal Component Analysis

A more conservative differentiation between aquatic and amphibious animals is used in the PCA. Any animal that has specialized morphology for an aquatic lifestyle and forages in an aquatic environment but can lift its body up off the ground when locomoting on land will be categorized as amphibious. Aquatic animals are animals that have specialized morphological features associated with life in an aquatic environment and forage in an aquatic environment and cannot lift their bodies off the ground (i.e., they must drag themselves, or flop using body undulations) are aquatic.

Principal components analyses (PCAs) were run on two datasets, one including all fossil and extant elements that underwent an analysis of anisotropy (Figure 5.13), and one dataset that consisted exclusively of fossil and extant femora (Figure 5.14). A summary of the variance of the

PCAs is listed in Table 5.2, and the loadings in Table 5.3.

PCA including all elements (Figure 5.13)

PC1 explains 54.9% of the variance of the limb elements and is positively loaded by average cortical area fraction (Ct.Ar/Tt.Ar; 0.6314), bone volume fraction (BV/TV; 0.550) and mode of anisotropy (Ani.M; 0.521). PC2 explains 29.2% of the variance and is most heavily

143

Figure 5.15. PCA plots of all limb elements used. Brown corresponds to terrestrial animals, green to amphibious, blue to aquatic and purple to fossil tetrapod elements. Summary and loadings in table 1 and 2. loaded by proximodistal length of the element (PD; 0.835). Together PC1 and PC2 (Figure 5.13

A) explain 84.1% of the variance.

PC3 explains 13.3% of the variance in the sample and is positively loaded by Ani.M

(0.651) but negatively loaded by the bone volume fraction (BV/TV; -0.564). Together PC1 and

PC3 (Figure 5.13 B) explain 68.2% of the variance and the components are loaded by variables other than size.

Table 5.2. Summary of Primary Component Analyses

PCA All Fossils

PC Eigenvalue % variance

1 2.19737 54.934

2 1.166 29.15

3 0.531807 13.295

4 0.104823 2.6206

144

PCA Femora Only

PC Eigenvalue % variance

1 2.05001 51.25

2 1.20199 30.05

3 0.736919 18.423

4 0.011078 0.27695

Table 5.3. PCA loadings

PCA all fossils

PC 1 PC 2 PC 3 PC 4

BV/TV 0.55001 0.34533 -0.56357 -0.51052

CT.Ar/Tt.Ar 0.6314 -0.21348 -0.17176 0.72544

Ani.M 0.52103 -0.37098 0.6512 -0.40847

PD (cm) 0.1654 0.83519 0.47837 0.21508

PCA femora only

PC 1 PC 2 PC 3 PC 4

BV/TV 0.52702 0.45966 -0.48536 0.52478

Ct.Ar/Tt.Ar 0.6836 -0.13307 -0.1439 -0.70305

Ani.M 0.50387 -0.35236 0.66751 0.41999

PD length (cm) 0.032657 0.80427 0.54603 -0.23224

145

Figure 5.16. PCA plots of femora only. Brown corresponds to terrestrial animals, green to amphibious, blue to aquatic and purple to fossil tetrapod elements. Summary and loadings in table 1 and 2. PCA of femora only (Figure 5.14)

PC1 explains 51.3% of the variance occurring in just the femora, and is almost primarily loaded by a combination of Ct.Ar/Tt.Ar (0.684), BV/TV (0.527), and Ani.M (0.504). PC2 explains 30.1% of the variance and is heavily influenced by PD (0.804). Together PC1 and PC2

(Figure 5.14 A) explain 81.4% of the variance within the data. However, PC2 is heavily loaded by size (PD) so the I will move onto the next component. PC3 explains 18.4% of the variance in the elements and is largely loaded by the median anisotropy value (0.668), and length (PD;

0.546). Together PC1 and PC3 (Figure 5.14 B) explain 67.7% of the variance in the fossil femora variables.

An additional PCA of the femora incorporating the four main compactness variables from bone profiler (Girondot and Laurin, 2003) is included in Appendix F.

146

Discussion

Femora

In the salamander Cryptobranchus (Figure 5.6 A-E) it was unclear if the structure within the bone was trabecular structure or another high-density material. The vector map and histogram are included in the figures, but not used in comparative analysis.

The extant taxa Uromastyx (Figure 5.4 F-J) and Felis (Figure 5.4 K-O) have vectors aligning in a predictable pattern at the proximal and distal ends of the bones. The patterns are consistent with what I predicted would occur from gravitational and ground reaction forces being transmitted through the pelvis and back up from the tibia and fibula respectively. These patterns are clearly observed for vectors with a degree of anisotropy 0.65 or higher (i.e. highly anisotropic). Eublepharis (Figure 5.4 A-E) has a less visible pattern of alignment to the regions of high anisotropy, but they are still present. Eublepharis, because of its size, may have not have as extensive remodelling. The force of gravity is acting on a smaller mass compared to the other terrestrial and amphibious animals examined. Therefore, there may be a size threshold, a minimum size for which there cease to be allometric effects. Additionally, the width of the femoral shaft of Eublepharis is only approximately 2mm (but trabecular thickness was comparable to the bigger Uromastyx) which also made it necessary but difficult to use a 500µm sampling and spacing parameter.

All three terrestrial femora have frequency of anisotropy histograms which have peaks

0.5 of higher (i.e. there is the greatest frequency of anisotropy values 0.5 or higher; Figure 5.4 E,

J, O). If the histograms do not have a clear peak, they do appear to have an upward slope as opposed to the aquatic taxa which all have anisotropy histograms with a downward slope, no

147

slope, or peak below 0.5 (Figure 5.6 E, J, O). Importantly, there are no clear patterns of internal high anisotropy vector orientation at the proximal and distal ends of the extant aquatic bones

(Figure 5.6), which is consistent with my prediction of the animals not experiencing gravitation and ground reaction forced through the pelvis and tibia/fibula respectively. Ornithyrynchus

(Figure 5.6 F-I) was an exception to this trend and is discussed below.

The highly aligned high anisotropy vectors located at the periphery of the trabecular bone along the trochanters of the platypus femur were unexpected, but descriptions of a dissection done on the hindlimb of a platypus (Gambaryan et al., 2002) show that these superficial but highly anisotropic regions correspond well with the attachment of the gluteus medius, vastus lateralis, and iliopsoas muscles (Figure 5.15). This anisotropic alignment appears to be the result of remodelling of trabecular bone associated with musculature rather than gravitational forces. I would expect more internal alignment and alignment within the femoral neck and head as well as the distal end if the trabeculae were responding to gravity. All three muscles that are exerting enough force for remodelling are muscles involved in flexing the hip and knee and rotating the limb in ways unique to the swimming style of Ornithorhynchus (Gambaryan et al., 2002).

Amblyrhynchus (Figure 5.6 K-O) also has highly anisotropic vectors that are located at the attachment of the PIFI (Molnar et al., 2019; figure 2 u). In Iguana the PIFI functions primarily to protract the femur (Russell and Bauer, 2008). Amblyrhynchus and Ornithirhynchus both have muscles involved with protracting and flexing the hip joint acting on the femora with enough force to induce remodelling. The trabecular remodelling of the aquatic (or amphibious, by strict definition) animals is not as extensive or aligned as in the terrestrial animals, which is consistent with my prediction. This shows that limb bones of animals with limited terrestrial locomotive behaviour undergo remodelling at muscle attachment sites as opposed to across joints. None of

148

the fossil femora had clear alignment of high anisotropy vectors at potential muscle attachment points (anywhere along the shaft, or adductor blades and crests) or at joint articulation surfaces, suggesting there are not enough muscular or terrestrial forces acting on the bones to induce remodelling, but these areas were often highly worn by taphonomic forces. This information does not support any of these animals having terrestrial behaviour.

The fossil femora NSM.005.GF.045.048-A (Figure 5.9 A-E), and NSM.007.GF.004.630

(Figure 5.9 F-J) both have a number of vectors with high anisotropy located all around the perimeter of the trabecular structure. However, the lack of clear pattern to the orientation and frequency histograms, which are more consistent with extant aquatic animals, leads me to conclude they were aquatic animals who did not use their limbs for terrestrial locomotion. The histograms are downward sloping with the highest frequencies of anisotropy occurring between

0.0 and 0.4. The distal femur NSM.005.GF.045.044 (Figure 5.10 A-E), which had the thickest trabeculae of the fossil femora, also is consistent with aquatic taxa.

An interesting detail of the frequency histograms is that all of the extant aquatic animals

(Figure 5.5 and 5.6), which are secondarily aquatic have distinct peaks at values of low anisotropy other than 0.0. Only two of the fossil elements share this configuration, the whatcheeriid-like distal femur NSM.005.GF.045.047 (Figure 5.10 J) and Crassigyrinus-like fibula NSM.005.GF.045.112 (Figure 5.11 E). As all extant aquatic animals are secondarily aquatic, and Crassigyrinus scoticus has been suggested as secondarily aquatic (Carroll, 1992;

Herbst and Hutchinson, 2019) I suggest this peak in anisotropy at a low value other than 0 may convey a signal about secondarily aquatic nature of the animal. The presence of secondarily aquatic animals at Blue Beach would suggest the presence of terrestrial tetrapods ancestral to the

149

Figure 5.17. Platypus femur. A, Muscle map adapted from Gambaryan et al (2002). Only the muscles corresponding to the aligned anisotropy vectors are illustrated here. B, vector map of the high anisotropy vectors (0.65 and above) of the whole femur. fauna found here. If this is the case the evolution of terrestriality would by necessity be pushed back into the Devonian, a concept supported by ichnofossil material (Stossel, 1995; Mossman and Grantham, 2008; Niedwiedzki et al., 2010). Alternatively, this anisotropy peak could be a signal associated with more amphibious animals as all the extant aquatic animals are capable of minimal locomotion on land (see strict definition of aquatic vs amphibious used for the PCAs below).

150

It should be noted that both of the fossil elements with low anisotropy peaks at values other than 0 are also cracked or partial elements. Therefore, it is possible that this is an artefact of the break in the fossil. Manually removing (slice-by-slice) erroneously labelled bone (i.e., where contrast between bone and matrix was indistinguishable from air and glue) might help reduce erroneous regions of high anisotropy due to cracks. However, some error will remain because of the edges created by the breaks in the trabecular bone.

The distal hindlimb fossil elements have more alignment to the high anisotropy vectors, often in the middle of the bone, near rugose regions. This could be from forces exerted on the bone by musculature similar to the extensor cruris/tarsi tibialis, flexor digitorum longus and quadratus femoris (Molnar et al. 2018) for an increased range of motion of the pes (Walthall and

Ashley-Ross, 2006). Molnar et al. (2017) observed that pectoral musculature of coelacanth and lungfish limbs often passes over the proximal element and attaches to more distal zeugopod and autopodial elements. It is also postulated that the hindlimb initially had delayed development of locomotor function compared to the forelimb (Pierce et al., 2012; Shubin et al., 2014), but others simply observe a lack of synchrony between the functional development of fore- and hindlimbs

(Coates et al., 2002). All the elements examined here, from a variety of contemporaneous animals, have distal elements of the hindlimb with more rugosity than the distal elements of the forelimb (ridges, pits, scars). None of the distal hindlimb elements have areas of highly patterned and oriented trabeculae at the proximal or distal ends, so it appears these limbs were not remodelling in response to gravity or ground reaction forces. A scenario that would explain the development of powerful hindlimbs that are not remodelling in response to gravity is that the animals from Blue Beach were using their hindlimbs for a more beneficial form of aquatic locomotion. This excludes animals such as Tiktaalik and Panderichthys from having moderately

151

weight-bearing forelimbs, but they are perhaps not as “ancestral” to other limbed tetrapods as previously thought (i.e. they are a distinct and distantly related lineage). Alternatively, the buoyancy of water sufficiently negates the force of gravity which is critical to secondary trabecular remodeling associated with terrestrial locomotion. Certainly, it appears that there was much more diversity in form, but perhaps not function, at Blue Beach.

Forelimb elements

Only one fossil tetrapod humerus was examined (Figure 5.8 A-E), as the others either have internal crushing, too many high-density inclusions, or were unavailable for study. The whatcheeriid humerus vector maps and anisotropy histogram are consistent with an aquatic limb.

The ulna (Figure 5.8 F-J) and radius (Figure 5.8 K-L) both have anisotropy histograms consistent with the frequency histograms of proximal elements of aquatic animals, and perhaps some detectable regions of high anisotropy patterns near soft tissue interaction sites; this requires further detailed investigation. The ulna and radius lack patterned areas of high anisotropy at the proximal and distal ends of the bones as would be expected with transsmission of gravitational forces and ground reaction forces if the animals were walking in a terrestrial environment.

A fossil humerus of the aquatic fish Eusthenopteron (Figure 5.7) is very similar to the

Blue Beach tetrapod elements. It has high anisotropy vectors around the periphery of the trabecular network, which are not clearly aligned. Anisotropy frequencies are consistent with

Blue Beach material. There is a very slight peak in frequencies of anisotropy around 0.3, but this is not as distinct as in the extant aquatic taxa or Blue Beach fossils with peaks in their anisotropy frequencies.

152

General observations

Trabecular organization in the extant animals was most obvious in the proximal and distal ends of the terrestrial animals, which was also where trabecular structure was almost exclusively located. This was consistent with predictions of how terrestrial trabeculae were expected to remodel. Aquatic extant animals had much more trabecular bone present throughout the long bone including at the midshaft, which was where most of the disorganized areas of high anisotropy occurred. Trabecular organization appeared internally in all the extant bones as opposed to being concentrated along the periphery of the trabecular structure as it did in the fossil elements. I observed that in most of the fossil specimens the trabecular bone thickened until it became dense cortical bone (making them difficult to separate), which led to poorly aligned high anisotropy vectors mapped at the periphery of the trabecular network. However, the humerus of Eusthenopteron, which had clear trabecular and cortical layers, also had a similar pattern of poorly aligned regions of anisotropy at the periphery of the trabecular structure in the vector map. This pattern did not appear in extant material or was less noticeable because trabecular orientation within the bone was also present.

Principal Component Analysis

PCA of all fossils

A plot of PC1 by PC2 (Figure 5.13 A) separates locomotor behaviours out into clusters with some overlap between terrestrial and amphibious animals. Extant aquatic animals and fossil taxa cluster together with no overlap between terrestrial and amphibious clusters. The cat femur anomalously groups with the larger amphibious animals and is responsible for the overlap between the amphibious and terrestrial clusters. The cat femur is from a juvenile animal

(demonstrated by the unossified epiphysis) and trabecular remodelling is expressed differently

153

between juvenile and adult animals (Tanck et al. 2001; Kivell, 2016). BV/TV changes occur rapidly in younger animals and measurable trabecular orientation responses occur later (Tanck et al., 2001). For a PCA excluding the cat see Appendix G. Size (PD) explains nearly all of the variance in PC2; however, this is due to a small sample size of elements where many of the taxa representing the different lifestyles by chance were of similar sizes (i.e., the size of the element is not informative about lifestyle).

On a plot of PC1 by PC3 (Figure 5.13 B), which explains 68.2% of the variance, and is less influenced by size, there is minimal overlap between locomotor behaviour clusters, except for the overlap caused by the Felis femur. Again, the fossil taxa cluster with the extant aquatic taxa and are distinct from both extant amphibious and extant terrestrial clusters.

PCA of Femora only

A plot of PC1 by PC2 (Figure 5.4 A) explains 81.3% of the total variance and does not create discrete amphibious and terrestrial clusters. As in the PCA of all the fossil elements terrestrial and amphibious animals overlap more with each other and are separate from the fossil animals. The turtle Elseya plots closer to the clustering of extant animals than extinct animals, but it is the closest of the extant animals to the fossil femora. The influence of size, which is not related to locomotor behaviour, on PC2 suggest an examination of PC1 By PC3 may be more successful to separate discrete groupings. The plot of PC1 by PC3 (Figure 5.14 B) explains

69.7% of the variance. This plot again shows that extant taxa cluster together apart from extinct femora. There is overlap between the terrestrial and amphibious taxa; however, Elseya, the aquatic turtle, separates out from the amphibious and aquatic taxa. The fossil femora do not cluster close to Elseya along the PC3 axis but along PC1 Elseya is the closest point to any of the fossil femora. A single data point representing an aquatic femur is not enough to make any

154

substantial conclusions with, as I cannot establish any potential range or overlap with respect to the fossil femora or extant terrestrial and amphibious femora.

Preliminary results suggest this method of evaluating the structural alignment of trabecular bone across the whole bone, and combining it with other whole bone measurements of

BV/TV and Cr.Ar/Tt.Ar provide a good method for inferring locomotor capabilities of animals across the fin-to-limb transition. Conclusions are drawn from physical features preserved in the bone, rather than features proposed from inferences as done when using the extant phylogenetic bracket (Bryant and Russell, 1992; Witmer, 1995) to reconstruct soft tissue (Molnar et al., 2017,

2018). The clustering of extant animals apart from extinct animals in the PCAs may suggest that bone development of early tetrapods occurred in a manner completely different than any modern tetrapod. Sanchez et al. (2016) proposed a unique method of development in the humerus of

Acanthostega, where an almost fully formed cartilaginous bone is ossified late in a long-lasting juvenile stage. Xradia (Zeiss, Germany) microCT scans were unable to image the difference between calcified cartilage and bone, but pits extending through the cortical layer of the Blue

Beach fossils would support the idea (although no specific mechanism is known) that patterns of ossification of early tetrapods are unlike those of modern tetrapods. Histological studies using higher resolution scans have the potential to reveal more of the three-dimensional internal microstructure and morphology of early tetrapod bone. This would enable inferences of physiology or the development of the bone, which could shed further light on the matter of bone remodelling in early tetrapods.

155

Conclusions

Based on the whole bone analysis of frequency of trabecular anisotropy, bone volume fraction, and cortical area fraction the Blue Beach tetrapods are more similar to aquatic than amphibious or terrestrial animals. This is despite the presence of morphotypes like later

Carboniferous animals suggested to be more capable of terrestrial locomotion (i.e., Pederpes).

None have clear patterns in alignment of high anisotropy vectors as observed in the extant terrestrial animals Eublepharis, Uromastyx, or Felis, and none have increased frequencies of high anisotropy like those of the terrestrial animals. Thus, in both respects they are more like the extant aquatic limb bones, which have unaligned vectors of high anisotropy and high frequency of low values of anisotropy. The PCA analysis of all elements support the conclusion that Blue

Beach tetrapods are most similar to aquatic animals with the provisos associated with low sample size. The PCA of femora only clearly separates fossil femora from extant femora, rather than locomotor behaviours. A larger sample of comparative taxa would help discriminate if these variables distinguish between locomotor groups or express a deeper difference between fossil limb bones and modern limb bones.

The possibility of secondarily aquatic taxa present in the early Tournaisian would extend the evolution of terrestrial tetrapods back into the Devonian, a concept supported by fossil trackways at Blue Beach (Mossman and Grantham, 2008; Mansky and Lucas, 2013), Poland

(Niedwiedzki et al., 2010), and Ireland (Stossel, 1995). These trackways preceding known walking animals present an interesting dilemma. Contemporaneous and younger animals with body fossils do not have features reflective of terrestrial locomotion. Questions remain about how the trackmakers were walking, through air or in water, and importantly why are there no body fossils of these animals.

156

The presence of aquatic taxa represented by body fossils at Blue Beach suggests there was a diverse faunal community with radiating body forms in an aquatic environment. This confounds the simple narrative of fish developing limb-like fins which were exapted for use on land and then radiated on land for various terrestrial locomotor strategies. It is now clear that tetrapods and their limbs radiated in an aquatic environment. The driving biotic and abiotic forces behind this aquatic diversification and radiation remains unclear, as does the unique physiology of early tetrapod limb bones. Both topics would be excellent avenues for further research.

This is the first study that examined anisotropy of early tetrapod limb bone material. A larger extant comparative sample size would make future studies more robust. This new method allows for subtle details to be picked up and trends across the whole bone to be considered. It also provides quantitative information that can be both qualitatively and quantitatively assessed regarding bone biological responses to locomotor behaviour.

Literature cited

Bennett, C. E., A. S. Howard, S. J. Davies, T. I. Kearsey, D. Millward, P. J. Brand, M. A. E.

Browne, E. J. Reeves, and J. E. A. Marshall. 2017. Ichnofauna record cryptic marine

incursions onto a coastal floodplain at a key Lower Mississippian tetrapod site.

Palaeogeography, Palaeoclimatology, Palaeoecology 468:287–300.

Biewener, A. A., N. L. Fazzalari, D. D. Konieczynski, and R. V Baudinette. 1996. Adaptive

changes in trabecular architecture in relation to functional strain patterns and disuse. Bone

19:1–8.

157

Bishop, P. J., S. A. Hocknull, C. J. Clemente, J. R. Hutchinson, A. A. Farke, B. R. Beck, R. S.

Barrett, and D. G. Lloyd. 2018. Cancellous bone and theropod dinosaur locomotion. Part

I—an examination of cancellous bone architecture in the hindlimb bones of theropods.

PeerJ 6:e5778.

Bryant, H. N., and K. L. Seymour. 1990. Observations and comments on the reliability of muscle

reconstruction in fossil vertebrates. Journal of Morphology 206:109–117.

Bryant, H. N., and A. P. Russell. 1992. The role of phylogenetic analysis in the inference of

unpreserved attributes of extinct taxa. Philosophical Transactions - Royal Society of

London, B 337:405–418.

Buie, H. R., G. M. Campbell, R. J. Klinck, J. A. MacNeil, and S. K. Boyd. 2007. Automatic

segmentation of cortical and trabecular compartments based on a dual threshold technique

for in vivo micro-CT bone analysis. Bone 41:505–515.

Canoville, A., and M. Laurin. 2009. Microanatomical diversity of the humerus and lifestyle in

lissamphibians. Acta Zoologica 90:110–122.

Carlson, K. J., S. Lublinsky, and S. Judex. 2008. Do different locomotor modes during growth

modulate trabecular architecture in the murine hind limb? Integrative and Comparative

Biology 48:385–393.

Carrano, M. T., and J. R. Hutchinson. 2002. Pelvic and hindlimb musculature of Tyrannosaurus

rex (Dinosauria: Theropoda). Journal of Morphology 253:207–228.

Carroll, R. L. 1992. The Primary Radiation of Terrestrial Vertebrates. Annual Review of Earth

and Planetary Sciences 20:45–84.

158

Clack, J. A. 2001. Eucritta melanolimnetes from the Early Carboniferous of Scotland, a stem

tetrapod showing a mosaic of characteristics. Transactions of the Royal Society of

Edinburgh: Earth Sciences 92:75–95.

Clack, J. A., C. E. Bennett, D. K. Carpenter, S. J. Davies, N. C. Fraser, T. I. Kearsey, J. E. A.

Marshall, D. Millward, B. K. A. Otoo, E. J. Reeves, A. J. Ross, M. Ruta, K. Z. Smithson,

T. R. Smithson, and S. A. Walsh. 2017. Phylogenetic and environmental context of a

Tournaisian tetrapod fauna. Nature Ecology & Evolution 1.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Coates, M. I., and J. A. Clack. 1990. Polydactyly in the earliest known tetrapod limbs. Nature

347:66–69.

Coates, M. I., J. E. Jeffery, and M. Ruta. 2002. Fins to limbs: what the fossils say. Evolution and

Development 4:390–401.

Coates, M. I., M. Ruta, and M. Friedman. 2008. Ever Since Owen: Changing Perspectives on the

Early Evolution of Tetrapods. Annual Review of Ecology, Evolution, and Systematics

39:571–592.

Cowin, S. C. 2001. The false premise in Wolff’s law; pp. 30-1-30–16 in Bone Mechanics

Handbook, Second Edition.

Diogo, R., P. Johnston, J. L. Molnar, and B. Esteve-Altava. 2016. Characteristic tetrapod

musculoskeletal limb phenotype emerged more than 400 MYA in basal lobe-finned fishes.

Scientific Reports 6:37592.

159

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

Dunmore, C. J., T. L. Kivell, A. Bardo, and M. M. Skinner. 2019. Metacarpal trabecular bone

varies with distinct hand-positions used in hominid locomotion. Journal of Anatomy

235:45–66.

Eurell, J., and D. C. van Sickle. 1998. Connective and Supportive Tissues; pp. 32–61 in H. D.

Dellmann and J. Eurell (eds.), Textbook of Vererinary Histology, 5th ed. Lippincott

Williams and Wilking, Baltimore.

Fajardo, R. J., and R. Müller. 2001. Three-dimensional analysis of nonhuman primate trabecular

architecture using micro-computed tomography. American Journal of Physical

Anthropology 115:327–336.

Gambaryan, P. P., A. A. Aristov, J. M. Dixon, and G. Y. Zubtsova. 2002. Peculiarities of the

hind limb musculature in monotremes: an anatomical description and functional approach.

Russian Journal of Theriology 1:1–36.

Georgiou, L., T. L. Kivell, D. H. Pahr, L. T. Buck, and M. M. Skinner. 2019. Trabecular

architecture of the great ape and human femoral head. Journal of Anatomy 234:679–693.

Germain, D., and M. Laurin. 2005. Microanatomy of the radius and lifestyle in amniotes

(Vertebrata, Tetrapoda). Zoologica Scripta 34:335–350.

Girondot, M., and M. Laurin. 2003. Bone Profiler: A tool to quantify, model, and statistically

compare bone-section compactness profiles. Journal of Vertebrate Paleontology 23:458–

461.

160

Goldstein, S. A., L. S. Matthews, J. L. Kuhn, and S. J. Hollister. 1991. Trabecular bone

remodeling: An experimental model. Journal of Biomechanics 24:135–150.

Hall, B. K. 2005. Bones and Cartilage: Developmental and Evolutionary Skeletal Biology.

Elsevier, San Diago, 1–760 pp.

Hall, B. K. 2007. Fins into Limbs: Evolution, Development, and Transformation. University of

Chicago Press.

Hammer, Ø., Harper, D.A.T., Ryan, P.D. 2001. PAST: Paleontological statistics software

package for education and data analysis. Palaeontologia Electronica 4(1): 9pp.

http://palaeo-electronica.org/2001_1/past/issue1_01.htm

Herbst, E. C., and J. R. Hutchinson. 2019. New insights into the morphology of the

Carboniferous tetrapod Crassigyrinus scoticus from computed tomography. Earth and

Environmental Science Transactions of the Royal Society of Edinburgh 109:157–175.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

Sciences 306:431–524.

Hutchinson, J. R. 2001a. The evolution of pelvic osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:123–168.

Hutchinson, J. R. 2001b. The evolution of femoral osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:169–197.

161

Hutchinson, J. R. 2002. The evolution of hindlimb tendons and muscles on the line to crown-

group birds. Comparative Biochemistry and Physiology - A Molecular and Integrative

Physiology 133:1051–1086.

Hutchinson, J. R., and S. M. Gatesy. 2000. Adductors, abductors, and the evolution of archosaur

locomotion. Paleobiology 26:734–751.

Ketcham, R. A., and T. M. Ryan. 2004. Quantification and visualization of anisotropy in

trabecular bone. Journal of Microscopy 213:158–171.

Kivell, T. L. 2016. A review of trabecular bone functional adaptation: What have we learned

from trabecular analyses in extant hominoids and what can we apply to fossils? Journal of

Anatomy 228:569–594.

Kivell, T. L., R. Davenport, J. J. Hublin, J. F. Thackeray, and M. M. Skinner. 2018. Trabecular

architecture and joint loading of the proximal humerus in extant hominoids, Ateles, and

Australopithecus africanus. American Journal of Physical Anthropology 167:348–365.

Kohler, T., M. Stauber, L. R. Donahue, and R. Müller. 2007. Automated compartmental analysis

for high-throughput skeletal phenotyping in femora of genetic mouse models. Bone

41:659–667.

Kriloff, A., D. Germain, A. Canoville, P. Vincent, M. Sache, and M. Laurin. 2008. Evolution of

bone microanatomy of the tetrapod tibia and its use in palaeobiological inference. Journal

of Evolutionary Biology 21:807–826.

Laurin, M., M. Girondot, and M.-M. Loth. 2004. The evolution of long bone microstructure and

lifestyle in lissamphibians. Paleobiology 30:589–613.

162

Laurin, M., A. Canoville, and D. Germain. 2011. Bone microanatomy and lifestyle: A

descriptive approach. Comptes Rendus - Palevol 10:381–402.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, A. T., D. C. McGregor, and J. Utting. 1993. Stratigraphic significance of Upper

Devonian and Lower Carboniferous miospores from the type area of the Horton Group,

Nova Scotia. Canadian Journal of Earth Sciences 30:1091–1098.

Martel, T. A., and M. R. Gibling. 1996. Stratigraphy and tectonic history of the Upper Devonian

to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology 32:13–38.

Meunier, F. J., and M. Laurin. 2012. A microanatomical and histological study of the fin long

bones of the Devonian sarcopterygian Eusthenopteron foordi. Acta Zoologica 93:88–97. von Meyer GH (1867) Die architektur der spongiosa. Arch Anat Physiol Wiss Med 34, 615–628.

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2017. Reconstructing pectoral

appendicular muscle anatomy in fossil fish and tetrapods over the fins-to-limbs transition.

Biological Reviews. 93:1077-1107

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2018. Evolution of Hindlimb Muscle

Anatomy Across the Tetrapod Water-to-Land Transition, Including Comparisons with

Forelimb Anatomy. Anatomical Record. doi: 10.1002/ar.23997.

Mossman, D. J., and R. G. Grantham. 2008. Eochelysipus horni, a new vertebrate trace fossil

from the Tournaisian Horton Bluff Formation, Nova Scotia. Atlantic Geology 44:69–77.

163

Nakajima, Y., and H. Endo. 2013. Comparative Humeral Microanatomy of Terrestrial,

Semiaquatic, and Aquatic Carnivorans using Micro-Focus CT Scan. Mammal Study 38:1–

8.

Niedwiedzki, G., P. Szrek, K. Narkiewicz, M. Narkiewicz, and P. E. Ahlberg. 2010. Tetrapod

trackways from the early Middle Devonian period of Poland. Nature 463:43–48.

Nyakatura, J. A., K. Melo, T. Horvat, K. Karakasiliotis, V. R. Allen, A. Andikfar, E. Andrada, P.

Arnold, J. Lauströer, J. R. Hutchinson, M. S. Fischer, and A. J. Ijspeert. 2019. Reverse-

engineering the locomotion of a stem amniote. Nature 565:351–355.

Otoo, B. K. A., J. A. Clack, T. R. Smithson, C. E. Bennett, T. I. Kearsey, and M. I. Coates. 2019.

A fish and tetrapod fauna from Romer’s Gap preserved in Scottish Tournaisian floodplain

deposits. Palaeontology 62:225–253.

Padian, K., and E.-T. Lamm. (Eds.). 2013. Bone Histology of Fossil Tetrapods: Advancing

Methods, Analysis, and Interpretation. University of California Press, Berkeley, California,

298 pp.

Petermann, H., and M. Sander. 2013. Histological evidence for muscle insertion in extant

amniote femora: Implications for muscle reconstruction in fossils. Journal of Anatomy

222:419–436.

Pierce, S. E., J. A. Clack, and J. R. Hutchinson. 2012. Three-dimensional limb joint mobility in

the early tetrapod Ichthyostega. Nature 486:523–526.

164

Plasse, M., E. Amson, J. Bardin, Q. Grimal, and D. Germain. 2019. Trabecular architecture in

the humeral metaphyses of non‐avian reptiles (Crocodylia, Squamata and Testudines):

Lifestyle, allometry and phylogeny. Journal of Morphology jmor.20996.

Quemeneur, S., V. de Buffrénil, and M. Laurin. 2013. Microanatomy of the amniote femur and

inference of lifestyle in limbed vertebrates. Biological Journal of the Linnean Society

109:644–655.

Reznikov, N., H. Chase, Y. Ben Zvi, V. Tarle, M. Singer, V. Brumfeld, R. Shahar, and S.

Weiner. 2016. Inter-trabecular angle: A parameter of trabecular bone architecture in the

human proximal femur that reveals underlying topological motifs. Acta Biomaterialia

44:65–72.

Roux W (1881) Der zu¨chtende Kampf der Teile, oder die ‘Teilauslee’im Organismus (Theorie

der ‘funktionellen Anpassung’). Leipzig: Wilhelm Engelmann

Russell, A. P., and A. M. Bauer. 2008. The appendicular locomotor apparatus of Sphenodon and

normal-limbed squamates. Biology of the Reptilia 21:1–465.

Ryan, T. M., and R. A. Ketcham. 2002a. The three-dimensional structure of trabecular bone in

the femoral head of strepsirrhine primates. Journal of Human Evolution 43:1–26.

Ryan, T. M., and R. A. Ketcham. 2002b. Femoral head trabecular bone structure in two

omomyid primates. Journal of Human Evolution 43:241–263.

Ryan, T. M., and R. A. Ketcham. 2005. Angular orientation of trabecular bone in the femoral

head and its relationship to hip joint loads in leaping primates. Journal of Morphology

265:249–263.

165

Sallan, L. C., M. I. Coates, J. Clack. 2010. End-Devonian extinction and a bottleneck in the early

evolution of modern jawed End-Devonian extinction and a bottleneck in the early

evolution of modern jawed vertebrates. Proceedings of the National Academy of Sciences

of the United States of America 107:10131–10135.

Sanchez, S., V. Dupret, P. Tafforeau, K. M. Trinajstic, B. Ryll, P. J. Gouttenoire, L. Wretman, L.

Zylberberg, F. Peyrin, and P. E. Ahlberg. 2013. 3D Microstructural Architecture of Muscle

Attachments in Extant and Fossil Vertebrates Revealed by Synchrotron Microtomography.

PLoS ONE 8:e56992.

Sanchez, S., P. Tafforeau, J. A. Clack, and P. E. Ahlberg. 2016. Life history of the stem tetrapod

Acanthostega revealed by synchrotron microtomography. Nature 537:408–411.

Sanchez, S., P. E. Ahlberg, K. M. Trinajstic, A. Mirone, and P. Tafforeau. 2012. Three-

dimensional synchrotron virtual paleohistology: A new insight into the world of fossil bone

microstructures. Microscopy and Microanalysis 18:1095–1105.

Sanchez, S., D. Germain, A. De Ricqlès, A. Abourachid, F. Goussard, and P. Tafforeau. 2010.

Limb-bone histology of temnospondyls: Implications for understanding the diversification

of palaeoecologies and patterns of locomotion of Permo-Triassic tetrapods. Journal of

Evolutionary Biology 23:2076–2090.

Shubin, N. H., E. B. Daeschler, and F. A. Jenkins. 2014. Pelvic girdle and fin of Tiktaalik roseae.

Proceedings of the National Academy of Sciences of the United States of America

111:893–899.

Smit, T. H., A. Odgaard, and E. Schneider. 1997. Structure and Function of Vertebral Trabecular

Bone. SPINE 22:2823–2833.

166

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Stossel, I. 1995. The discovery of a new Devonian tetrapod trackway in SW Ireland. Journal -

Geological Society (London) 152:407–413.

Su, A., and K. J. Carlson. 2017. Comparative analysis of trabecular bone structure and

orientation in South African hominin tali. Journal of Human Evolution 106:1–18.

Tanck, E., J. Homminga, G. H. Van Lenthe, and R. Huiskes. 2001. Increase in bone volume

fraction precedes architectural adaptation in growing bone. Bone 28:650–654.

Tibert, N. E., and D. B. Scott. 1999. Ostracodes and Agglutinated Foraminifera as Indicators of

Paleoenvironmental Change in an Early Carboniferous Brackish Bay, Atlantic Canada.

PALAIOS 14:246.

Tsegai, Z. J., M. M. Skinner, D. H. Pahr, J. J. Hublin, and T. L. Kivell. 2018. Systemic patterns

of trabecular bone across the human and chimpanzee skeleton. Journal of Anatomy

232:641–656.

Ulrich, D., B. Van Rietbergen, A. Laib, and P. R̈ uegsegger. 1999. The ability of three-

dimensional structural indices to reflect mechanical aspects of trabecular bone. Bone

25:55–60.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

167

Wallace, I. J., A. T. Kwaczala, S. Judex, B. Demes, and K. J. Carlson. 2013. Physical activity

engendering loads from diverse directions augments the growing skeleton. Journal of

Musculoskeletal & Neuronal Interactions 13:283–288.

Ward, P., C. Labandeira, M. Laurin, and R. A. Berner. 2006. Confirmation of Romer’s Gap as a

low oxygen interval constraining the timing of initial arthropod and vertebrate

terrestrialization. Proceedings of the National Academy of Sciences 103:16818–16822.

Warren, A. 2007. New data on Ossinodus pueri, a stem tetrapod from the Early Carboniferous of

Australia. Journal of Vertebrate Paleontology 27:850–862.

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Waters, C. N., R. A. Waters, W. J. Barclay, and J. R. Davies. 2009. A Lithostratigraphical

Framework for the Carboniferous Successions of Northern Great Britain (Onshore). 174

pp.

Williams-Hatala, E. M., K. G. Hatala, M. Gordon, A. Key, M. Kasper, and T. L. Kivell. 2018.

The manual pressures of stone tool behaviors and their implications for the evolution of the

human hand. Journal of Human Evolution 119:14–26.

Witmer, L. M. 1995. The Extant Phylogenetic Bracket and the importance of reconstructing soft

tissues in fossils; pp. 19–33 in Jeff Thomason (ed.), Functional morphology in vertebrate

paleontology. Cambridge University Press.

Wolff J (1892) Das Gesetz der Transformation der Knochen. Berlin: A. Hirchwild.

168

Zeininger, A., B. A. Patel, B. Zipfel, and K. J. Carlson. 2016. Trabecular architecture in the StW

352 fossil hominin calcaneus. Journal of Human Evolution 97:145–158.

169

Chapter 6: Summary and Conclusions The purpose of this research project was to test and apply methods relying on both external and internal data of early tetrapod bone to assess locomotor behaviours of the Blue

Beach fauna. Proxies based on physiological and biological processes may provide a more robust signal than inferences based on external morphological features. An additional new method was proposed and qualitatively and quantitatively tested using whole-bone data as opposed to regional data.

Chapter 1 summary

Chapter One introduced the history of problems associated with study of the fin-to-limb and water-to-land transitions. These include biases towards complete articulated material that resulted in a temporary lack of known material and proxies for locomotion that are subjective and lack rigorous testing. I then outlined how, after describing the previously undescribed Blue

Beach material borrowed from the BBFM, I would use external features including osteological correlates to assess in which direction muscles were acting on the femora. Additionally, I proposed using internal quantitative proxies including a method and formula previously used by

Quemeneur et al (2013) to infer terrestrial capability of a range of amniote femora. Finally, I proposed a new method to qualitatively assess whole three-dimensional bone volume features associated with bone functional adaptation responses.

Chapter 2 summary

The second chapter described a range of limb material from early tetrapods including animals belonging to Devonian and Carboniferous morphotypes. These morphotypes are often associated with aquatic (Acanthostegid and Tulerpetontid; Lebedev and Coates, 1995; Coates,

170

1996) to amphibious (Holmes, 1984; Smithson, 1985; Clack and Finney, 2005) locomotor behaviours based on external features such as symmetry and dactyly of the manus and pes (hand and foot), as well as orientation, ossification, and robusticity of the limb bones. Much of the fossil material from Blue Beach had a mosaic of features known from described fossil tetrapods.

No animals were described beyond a genus-like morphotype. Much of the material preserved rugosity which is important for attempting soft tissue inferences and analysis.

Chapter 3 summary

In the third chapter I attempted to infer locomotory behaviours in fossil limb bones from osteological correlates of soft tissue attachment. I used extant animals to determine direction and differentiations of musculature from osteological correlates of two reptiles and two amphibians.

The animals included aquatic (Necturus), amphibious (Cynops) and terrestrial (Eublepharis and

Uromastyx) locomotor behaviours. Unfortunately, due to the inconsistent nature of muscle attachments (highly variable topology in extant tetrapods, and a large proportion that do not leave traces on bone) reconstruction of muscle maps showing location and direction of attachment of musculature was not possible. Data from osseous material only was explicitly used once it became clear that contrast enhanced microCT reconstruction of musculature led to problems beyond the scope of the project (i.e., failure to visualize tendon and inconsistencies between virtual and manual dissections and descriptions in the literature). The identification of topological similarities between ridges along the shaft of Cynops, and the adductor crest and posterior edge of the popliteal area of fossil femur YPM.PU.20103 (but a consistent feature in all the Blue Beach femora) which are in locations consistent with muscles involved in elevating, protracting and retracting the femur (PIFE and PIFI respectively). This suggests muscle were pulling on the femur to rotate the limb more. Another discovery of this study was the

171

identification of a misconception regarding Sharpey’s and muscle insertion fibers preserved on fossil bones. Pits too large to be remnants of microscopic muscle attachment fibers have been misinterpreted as correlates of such fibers. This has led to misinformed interpretations of musculature location, directionality, and possibly function. Past and future inferences based on identification of preservation of muscle insertion fibers should be reassessed. Clarifying vague terminology such as “scar”, “rugosity”, “pit”, and “enthesis” with more emphasis on synonymizing and understanding terms from literature on extant animals and biomedical literature will help make inferences about such bone features more consistent and reliable. The presence of unusual pits previously misinterpreted and muscle attachment fibers and the observation that they extend through the entire cortical layer of bone suggests that early tetrapods had an extensive network of nutrient transport or innervation extending into the limb bones. Such extensive pitting is not a feature seen in modern tetrapods. This may be an additional line of evidence that early tetrapod bone underwent physiological or developmental processes that are simply not comparable with modern analogues (Sanchez et al., 2016). Additionally, the study highlighted how variable musculature and bone morphology is within a single group of extant tetrapods and it illustrates that there is no linear gradation of aquatic to amphibious to terrestrial tetrapod femora.

Chapter 4 summary

In the fourth chapter I applied a previously published formula (Quemeneur et al., 2013) to resultant variables from a cross section analysis of early tetrapod femora. Bone Profiler

Software (Girondot and Laurin, 2003) was used to collect compactness profiles of the midshafts of Blue Beach femora. The variable Rmin, which was capable of predicting terrestrial/amphibious or aquatic locomotor behaviour was run through the binary formula provided by Quemeneur et

172

al. (2013). Only femora were used because they were the most abundant and had the most circular midshafts of all the limb elements.

The fossil femora included animals similar to aquatic Tulerpeton, and those believed to be more terrestrial such as Pederpes. The data Quemeneur et al. (2013) used midshafts of 155 species of extant amniotes to create their binary formula. The use of this many animals provides greater quantitative robusticity (than soft tissue inferences) to their results and therefore my results produced using their formula. All Blue Beach femora had compactness values that were non-zero at the bone centre in contrast to the terrestrial comparative sample. The fossil femora are most similar to the extant comparative aquatic animals, but exhibit deviations from both terrestrial representative Felis and aquatic representative Ornithorhynchus used to confirm the accuracy of the formula provided by Quemeneur et al (2013). Differences observed between graphical representations of the profiles within the Blue Beach fossil femora are also interesting.

They could suggest subtle differences in type of aquatic locomotion used (kicking, gliding, paddling), or primary or secondary aquatic nature of the limb. Perhaps the difference between the extant terrestrial and aquatic taxa versus the Blue Beach femora may indicate a deeper fundamental physiological or developmental difference between extinct and extant animals. It could be that early tetrapods were so physiologically divergent to any extant animals that none provide a good basis of comparison. This would be consistent with what was observed in the second chapter where pits that were previously interpreted as muscle attachment correlates are now suggested to be associated with vasculature or neurovasculature.

The results of Chapter Four emphasize that internal features have the potential to preserve more subtle differences in bone physiology and by extension inferences of function than those from external morphology. The midshaft data from isolated limb bones may not produce

173

fine discriminant results but there is enough information to determine that the Blue Beach animals are more consistent with extant aquatic tetrapods than amphibious or terrestrial ones.

Use of a greater amount of internal and three-dimensional data was used in the following chapter where numerous small volumes across the entire bone were used to produce quantifiable variables and results to infer locomotor behaviours.

Chapter 5 summary

The fifth chapter of this study involved developing a new method modified from previous methods used in the study of mammalian fossil material and concepts well known in medical research: anisotropy and bone functional adaptation. Using Dragonfly’s (ORS, Montréal) Bone

Analysis program, small regions (volumes) of interest were examined to determine the direction of segments of trabecular bone across an entire trabecular network. Patterns in the alignment of regions with high anisotropy (organization) were clearly visible in extant terrestrial animals, especially at the proximal and distal ends. This is consistent with where the force of gravity and ground reaction forces are transmitted across joints. Aquatic animals lacked highly anisotropic areas in the proximal and distal regions, and only the platypus Ornithorhynchus showed patterned regions of anisotropy. The patterned anisotropy in Ornithorhynchus is superficial (at the periphery of the trabecular network) and matches well with musculature acting on the region

(gluteus medius, vastus lateralis, and iliopsoas). This suggests that vector maps of limb bones may potentially be informative of musculature on amphibious to terrestrial animals when surface features do not provide osteological correlates. The highly anisotropic areas of the fossil femora were located exclusively along the periphery of the trabecular structure (i.e., where the delineation between cortical bone and trabecular bone occurred during segmentation). This is likely a factor of segmentation and does not convey a biological signal in terms of orientation of

174

the trabecular structure, but it does indicate that the trabecular bone thickens in such a way that as it approaches the cortical bone, the two are difficult to segment from each other. This again makes the fossil femora dissimilar to both extant aquatic and terrestrial animals used for comparison. Further support is provided by the scan of a humerus of Eusthenopteron, an animal that is indisputably aquatic, and is more similar to the other fossil limb bones (thin cortical structure and trabecular bone throughout the whole marrow cavity) than any of the extant taxa.

The lack of patterned alignment to trabecular anisotropy vectors of the Blue Beach femora, especially at the proximal and distal ends is more consistent with extant aquatic animals than terrestrial animals. It suggests that the fossil animals were not remodelling their bones in response to gravity. The more distal elements display similar results, where there is no pattern to the alignment of regions of high anisotropy, but there are many regions that are highly anisotropic. The few forelimb elements (including a Pederpes-like humerus) that were preserved well enough in 3D for trabecular analysis also have no pattern to the alignment of regions of high anisotropy.

Ratios of BV/TV are suggested to be informative about forces acting on a bone in conjunction with anisotropy in previous studies (Ryan and Ketcham, 2002a, 2002b; Zeininger et al., 2016; Su and Carlson, 2017; Plasse et al., 2019). In the analyses carried out in Chapter Four I found the BV/TV, in conjunction with mode of frequency of anisotropy and cortical area fraction across the entire bone, are useful in inferring locomotor behaviour of the fossil limb bone.

Qualitatively, all terrestrial extant taxa displayed histograms where instances of high anisotropy were more common than instances of low anisotropy when anisotropic values of 0.0 are discounted. In contrast, instances of low anisotropy were observed more often in aquatic taxa.

Blue Beach animals have even greater frequencies of lower anisotropy values than the extant

175

aquatic animals sampled. Whether this is a factor of differentiation between amphibious and aquatic, or primarily and secondarily aquatic limb bones is unclear and may be difficult to determine as there are no extant primarily aquatic tetrapods. A PCA of all the fossil elements quantitatively separated out fairly discrete locomotor behaviour functional groups. The fossil femora were clearly separate from terrestrial and amphibious animals and overlapped with the two known aquatic limb bones.

Femora were the most common and best preserved internally of all the Blue Beach material used for this study, as a result they were preferentially used over humeri and more distal elements. Further studies into those forelimb and distal elements provide a starting point for further studies.

Concluding Remarks

Descriptions of a variety of Devonian and Carbonifeorus morphotypes are consistent with previous analysis of the site (Mansky and Lucas, 2013; Anderson et al., 2015). However, the variety of morphotypes does not appear to reflect a variety of locomotor behaviours of the animals they belonged to. The combined results of Chapters Four and Five indicate that all of the animals from Blue Beach show more similarities with extant aquatic than terrestrial tetrapods, including animals that were expected to be consistent with more amphibious to terrestrial morphotypes. The results from these chapters most clearly show that early tetrapod limb bones are more similar to each other than they are to any extant aquatic or terrestrial animals used in this analysis. I suggest that this is related to the secondarily aquatic nature of all known extant aquatic tetrapods. Furthermore, observations of extensive perforations through the cortical bone on all of the proximal hind limb fossil material suggest that there are physiological processes occurring that are very different from any known in extant animals.

176

Despite the unreliability of muscle attachment inferences and by extension analyses based on muscle attachment inferences, fossils preserve other features that may be used to infer accurate information about biomechanical responses to gravity and thus locomotor behaviour.

Using internal anatomy is also a more direct way to infer function without relying on an extant phylogenetic bracket which results in an inference, to use as data for further inferences. The two- dimensional internal data from midshaft compactness profiles predict the various morphotypes of

Blue Beach femora to be aquatic rather than amphibious/terrestrial tetrapods. Incorporation of three-dimensional trabecular data show elements that lack patterned alignment to the high anisotropy regions, which is consistent with the amphibious and aquatic comparative sample.

More quantitative analyses of the three-dimensional data result in frequency histograms which display more frequent occurrences of regions of low anisotropy, and clustering of fossil limb bones with known aquatic limb bones in a PCA. Together the methods using internal limb bone data to infer locomotory behaviour of this multimethod approach suggest that the animals preserved within the diverse Blue Beach fauna were aquatic.

Literature Cited

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

177

Coates, M. I. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: Postcranial anatomy,

basal tetrapod interrelationships and patterns of skeletal evolution. Transactions of the

Royal Society of Edinburgh, Earth Sciences 87:363–421.

Girondot, M., and M. Laurin. 2003. Bone Profiler: A tool to quantify, model, and statistically

compare bone-section compactness profiles. Journal of Vertebrate Paleontology 23:458–

461.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

Sciences 306:431–524.

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Plasse, M., E. Amson, J. Bardin, Q. Grimal, and D. Germain. 2019. Trabecular architecture in

the humeral metaphyses of non‐avian reptiles (Crocodylia, Squamata and Testudines):

Lifestyle, allometry and phylogeny. Journal of Morphology jmor.20996.

Quemeneur, S., V. de Buffrénil, and M. Laurin. 2013. Microanatomy of the amniote femur and

inference of lifestyle in limbed vertebrates. Biological Journal of the Linnean Society

109:644–655.

178

Ryan, T. M., and R. A. Ketcham. 2002a. The three-dimensional structure of trabecular bone in

the femoral head of strepsirrhine primates. Journal of Human Evolution 43:1–26.

Ryan, T. M., and R. A. Ketcham. 2002b. Femoral head trabecular bone structure in two

omomyid primates. Journal of Human Evolution 43:241–263.

Sanchez, S., P. Tafforeau, J. A. Clack, and P. E. Ahlberg. 2016. Life history of the stem tetrapod

Acanthostega revealed by synchrotron microtomography. Nature 537:408–411.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Su, A., and K. J. Carlson. 2017. Comparative analysis of trabecular bone structure and

orientation in South African hominin tali. Journal of Human Evolution 106:1–18.

Zeininger, A., B. A. Patel, B. Zipfel, and K. J. Carlson. 2016. Trabecular architecture in the StW

352 fossil hominin calcaneus. Journal of Human Evolution 97:145–158.

179

Literature Cited Abdala, V., and R. Diogo. 2010. Comparative anatomy, homologies and evolution of the pectoral

and forelimb musculature of tetrapods with special attention to extant limbed amphibians

and reptiles. Journal of Anatomy 217:536–573.

Anderson, J. S., T. Smithson, C. F. Mansky, T. Meyer, and J. Clack. 2015. A diverse tetrapod

fauna at the base of “Romer’s Gap.” PLoS ONE 10:1–27.

Barrell, J., E. Huntington, R. S. Lull, C. Schuchert, and L. L. Woodruff. 2011. The Evolution of

the Earth and Its Inhabitants; a Series of Lectures Delivered before the Yale Chapter of the

Sigma Xi during the Academic Year 1916-1917, by Joseph Barrell, Charles Schuchert,

Lorande Loss Woodruff, Richard Swann Lull, Ellsworth Huntingto, 3rd ed. Yale

University Press, New Haven, 1–235 pp.

Bennett, C. E., T. I. Kearsey, S. J. Davies, D. Millward, J. A. Clack, T. R. Smithson, and J. E. A.

Marshall. 2016. Early Mississippian sandy siltstones preserve rare vertebrate fossils in

seasonal flooding episodes. Sedimentology 63:1677–1700.

Bennett, C. E., A. S. Howard, S. J. Davies, T. I. Kearsey, D. Millward, P. J. Brand, M. A. E.

Browne, E. J. Reeves, and J. E. A. Marshall. 2017. Ichnofauna record cryptic marine

incursions onto a coastal floodplain at a key Lower Mississippian tetrapod site.

Palaeogeography, Palaeoclimatology, Palaeoecology 468:287–300.

Biewener, A. A., N. L. Fazzalari, D. D. Konieczynski, and R. V Baudinette. 1996. Adaptive

changes in trabecular architecture in relation to functional strain patterns and disuse. Bone

19:1–8.

180

Bishop, P. J. 2014. The humerus of Ossinodus pueri , a stem tetrapod from the Carboniferous of

Gondwana, and the early evolution of the tetrapod forelimb. Alcheringa: An Australasian

Journal of Palaeontology 38:209–238.

Bishop, P. J., S. A. Hocknull, C. J. Clemente, J. R. Hutchinson, A. A. Farke, B. R. Beck, R. S.

Barrett, and D. G. Lloyd. 2018. Cancellous bone and theropod dinosaur locomotion. Part

I—an examination of cancellous bone architecture in the hindlimb bones of theropods.

PeerJ 6:e5778.

Boisvert, C. A., E. Mark-Kurik, and P. E. Ahlberg. 2008. The pectoral fin of Panderichthys and

the origin of digits. Nature 456:636–638.

Bryant, H. N., and K. L. Seymour. 1990. Observations and comments on the reliability of muscle

reconstruction in fossil vertebrates. Journal of Morphology 206:109–117.

Bryant, H. N., and A. P. Russell. 1992. The role of phylogenetic analysis in the inference of

unpreserved attributes of extinct taxa. Philosophical Transactions Royal Society of

London, B 337:405–418.

Canoville, A., and M. Laurin. 2009. Microanatomical diversity of the humerus and lifestyle in

lissamphibians. Acta Zoologica 90:110–122.

Carroll, R. L., E. S. Belt, D. L. Dineley, D. Baird, and D. C. McGregor. 1972. International

Geological Congress: Twenty Fourth Session. 1–113 pp.

Carter, D. R., T. E. Orr, and D. P. Fyhrie. 1989. Relationships between loading history and

femoral cancellous bone architecture. Journal of Biomechanics 22:231–44.

181

Clack, J. A. 2001. Eucritta melanolimnetes from the Early Carboniferous of Scotland, a stem

tetrapod showing a mosaic of characteristics. Transactions of the Royal Society of

Edinburgh: Earth Sciences 92:75–95.

Clack, J. A. 2002. An early tetrapod from “Romer’s Gap”. Nature 418:72–6.

Clack, J. A. 2006. The emergence of early tetrapods. Palaeogeography, Palaeoclimatology,

Palaeoecology 232:167–189.

Clack, J. A., C. E. Bennett, D. K. Carpenter, S. J. Davies, N. C. Fraser, T. I. Kearsey, J. E. A.

Marshall, D. Millward, B. K. A. Otoo, E. J. Reeves, A. J. Ross, M. Ruta, K. Z. Smithson,

T. R. Smithson, and S. A. Walsh. 2017. Phylogenetic and environmental context of a

Tournaisian tetrapod fauna. Nature Ecology & Evolution 1.

Clack, J. A., and E. P. Ahlberg. 2004. A new stem tetrapod from the Early Carboniferous of

Northern Ireland. Recent Advances in the Origin and Early Radiation of Vertebrates 309–

320.

Clack, J. A., and S. M. Finney. 2005. Pederpes finneyae, an articulated tetrapod from the

Tournaisian of Western Scotland. Journal of Systematic Palaeontology 2:311–346.

Clack, J. A., R. L. Carroll, and H. Heatwole. 2000. Early Carboniferous tetrapods; pp. 1030–

1043 in Amphibian Biology. vol. 4 Palaeontology.

Clack, J. A., L. B. Porro, and C. E. Bennett. 2018. A Crassigyrinus-like jaw from the

Tournaisian (Early Mississippian) of Scotland. Earth and Environmental Science

Transactions of the Royal Society of Edinburgh 108:37–46.

182

Clack, J. A., C. E. Bennett, S. J. Davies, A. C. Scott, J. E. Sherwin, and T. R. Smithson. 2019. A

Tournaisian (earliest Carboniferous) conglomerate-preserved non-marine faunal

assemblage and its environmental and sedimentological context. PeerJ 2019:e5972.

Coates, M. I. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: Postcranial anatomy,

basal tetrapod interrelationships and patterns of skeletal evolution. Transactions of the

Royal Society of Edinburgh, Earth Sciences 87:363–421.

Coates, M. I., and J. A. Clack. 1990. Polydactyly in the earliest known tetrapod limbs. Nature

347:66–69.

Coates, M. I., and J. a. Clack. 1991. Fish-like gills and breathing in the earliest known tetrapod.

Nature 352:234–236.

Coates, M. I., and J. A. Clack. 1995. Romer’s gap: tetrapod origins and terrestriality. Bulletin Du

Muséum National d’Histoire Naturelle 4:373–388.

Coates, M. I., M. Ruta, and M. Friedman. 2008. Ever Since Owen: Changing Perspectives on the

Early Evolution of Tetrapods. Annual Review of Ecology, Evolution, and Systematics

39:571–592.

Cowin, S. C. 2001. The false premise in Wolff’s law; pp. 30-1-30–16 in Bone Mechanics

Handbook, Second Edition.

Craig, J. M., D. A. Mifsud, A. S. Briggs, J. Boase, and G. Kennedy. 2016. Mudpuppy (Necturus

maculosus maculosus) spatial distribution, breeding water depth, and use of artificial

spawning habitat in the Detroit river. Herpetological Conservation and Biology 10:926–

934.

183

Cunningham, P. 2000. Daily activity pattern and diet of a population of the spiny tailed lizard,

Uromastyx aegyptius microlepis, during summer in the United Arab Emirates. Zoology in

the Middle East 21:37–46.

Currey, J. D., and R. M. N. Alexander. 1985. The thickness of the walls of tubular bones. Journal

of Zoology 206:453–468.

Daeschler, E. B., N. H. Shubin, and F. A. Jenkins. 2006. A Devonian tetrapod-like fish and the

evolution of the tetrapod body plan. Nature 440:757–763.

Diogo, R., V. Abdala, N. Lonergan, and B. A. Wood. 2008. From fish to modern humans -

Comparative anatomy, homologies and evolution of the head and neck musculature.

Journal of Anatomy 213:391–424.

Diogo, R., P. Johnston, J. L. Molnar, and B. Esteve-Altava. 2016. Characteristic tetrapod

musculoskeletal limb phenotype emerged more than 400 MYA in basal lobe-finned fishes.

Scientific Reports 6:37592.

Diogo, R., S. Walsh, C. Smith, J. M. Ziermann, and V. Abdala. 2015. Towards the resolution of

a long-standing evolutionary question: Muscle identity and attachments are mainly related

to topological position and not to primordium or homeotic identity of digits. Journal of

Anatomy 226:523–529.

Dragonfly 4.0 [Computer software]. Object Research Systems (ORS) Inc, Montreal, Canada,

2019; software available at http://www.theobjects.com/dragonfly.

184

Dunmore, C. J., T. L. Kivell, A. Bardo, and M. M. Skinner. 2019. Metacarpal trabecular bone

varies with distinct hand-positions used in hominid locomotion. Journal of Anatomy

235:45–66.

Eurell, J., and D. C. van Sickle. 1998. Connective and Supportive Tissues; pp. 32–61 in H. D.

Dellmann and J. Eurell (eds.), Textbook of Vererinary Histology, 5th ed. Lippincott

Williams and Wilking, Baltimore.

Frost, H. M. 1990. Skeletal structural adaptations to mechanical usage (SATMU): 1. Redefining

Wolff’s Law: The bone modeling problem. The Anatomical Record 226:403–413.

Fuller, P. O., T. E. Higham, and A. J. Clark. 2011. Posture, speed, and habitat structure: Three-

dimensional hindlimb kinematics of two species of padless geckos. Zoology 114:104–112.

Germain, D., and M. Laurin. 2005. Microanatomy of the radius and lifestyle in amniotes

(Vertebrata, Tetrapoda). Zoologica Scripta 34:335–350.

Gignac, P. M., and A. Watanabe. 2016. Get Your Brights Brighter & Darks Darker: CLAHE +

diceCT – DiceCT. Available at https://dicect.com/2016/08/11/clahe/. Accessed August 26,

2019.

Gignac, P. M., N. J. Kley, J. A. Clarke, M. W. Colbert, A. C. Morhardt, D. Cerio, I. N. Cost, P.

G. Cox, J. D. Daza, C. M. Early, M. S. Echols, R. M. Henkelman, A. N. Herdina, C. M.

Holliday, Z. Li, K. Mahlow, S. Merchant, J. Müller, C. P. Orsbon, D. J. Paluh, M. L. Thies,

H. P. Tsai, and L. M. Witmer. 2016. Diffusible iodine-based contrast-enhanced computed

tomography (diceCT): An emerging tool for rapid, high-resolution, 3-D imaging of

metazoan soft tissues. Journal of Anatomy 228:889–909.

185

Girondot, M., and M. Laurin. 2003. Bone Profiler: A tool to quantify, model, and statistically

compare bone-section compactness profiles. Journal of Vertebrate Paleontology 23:458–

461.

Hall, B. K. 2005. Bones and Cartilage: Developmental and Evolutionary Skeletal Biology.

Elsevier, San Diago, 1–760 pp.

Hall, B. K. 2007. Fins into Limbs: Evolution, Development, and Transformation. University of

Chicago Press, 433 pp.

Holmes, R. 1984. The Carboniferous Amphibian Proterogyrinus scheelei Romer, and the Early

Evolution of Tetrapods. Philosophical Transactions of the Royal Society B: Biological

Sciences 306:431–524.

Hua, S., and V. De Buffrenil. 1996. Bone histology as a clue in the interpretation of functional

adaptations in the thalattosuchia (reptilia, crocodylia). Journal of Vertebrate Paleontology

16:703–717.

Huang, L.-K., and M.-J. Wang. 1995. Image thresholding by minimizing the measures of

fuzziness. Pattern Recognition 28:41–51.

Hutchinson, J. R. 2001. The evolution of pelvic osteology and soft tissues on the line to extant

birds (Neornithes). Zoological Journal of the Linnean Society 131:123–168.

Hutchinson, J. R. 2002. The evolution of hindlimb tendons and muscles on the line to crown-

group birds. Comparative Biochemistry and Physiology - A Molecular and Integrative

Physiology 133:1051–1086.

186

Hutchinson, J. R., and S. M. Gatesy. 2000. Adductors, abductors, and the evolution of archosaur

locomotion. Paleobiology 26:734–751.

Inger, R. F. 1957. Ecological Aspects of the Origins of the Tetrapods. Evolution 11:373.

Jarvik, E. 1996. The Devonian tetrapod Ichthyostega. Lethaia 29:76–76.

Jones, S. J., and A. Boyde. 1974. The organization and gross mineralization patterns of the

collagen fibres in sharpey fibre bone. Cell and Tissue Research 148:83–96.

Kapur, J. N., P. K. Sahoo, and A. K. C. Wong. 1985. A new method for gray-level picture

thresholding using the entropy of the histogram. Computer Vision, Graphics, & Image

Processing 29:273–285.

Kivell, T. L. 2016. A review of trabecular bone functional adaptation: What have we learned

from trabecular analyses in extant hominoids and what can we apply to fossils? Journal of

Anatomy 228:569–594.

Kriloff, A., D. Germain, A. Canoville, P. Vincent, M. Sache, and M. Laurin. 2008. Evolution of

bone microanatomy of the tetrapod tibia and its use in palaeobiological inference. Journal

of Evolutionary Biology 21:807–826.

Lanyon, L. E., A. E. Goodship, C. J. Pye, and J. H. MacFie. 1982. Mechanically adaptive bone

remodelling. Journal of Biomechanics 15:141–154.

Laurin, M., M. Girondot, and M.-M. Loth. 2004. The evolution of long bone microstructure and

lifestyle in lissamphibians. Paleobiology 30:589–613.

Laurin, M., A. Canoville, and D. Germain. 2011. Bone microanatomy and lifestyle: A

descriptive approach. Comptes Rendus - Palevol 10:381–402.

187

Lebedev, O. A., and M. I. Coates. 1995. The postcranial skeleton of the Devonian tetrapod

Tulerpeton curtum Lebedev. Zoological Journal of the Linnean Society 114:307–348.

Lieberman, D. E. 1997. Making Behavioral and Phylogenetic Inferences from Hominid Fossils:

Considering the Developmental Influence of Mechanical Forces. Annual Review of

Anthropology 26:185–210.

Lombard, E. F., and J. R. Bolt. 1995. A New Primitive Tetrapod, Whatcheeria Deltae, From The

Lower Carboniferous of Iowa. Palaeontology 38:471–494.

Long, J. A., and M. S. Gordon. 2004. The greatest step in vertebrate history: A paleobiological

review of the fish-tetrapod transition. Physiological and Biochemical Zoology 77:700–719.

Mansky, C. F., and S. G. Lucas. 2013. Romer’s Gap Revisited: Continental Assemblages And

Ichno-Assemblages From The Basal Carboniferous Of Blue Beach, Nova Scotia. New

Mexo Museum of Natural History and Science Bulletin 60:244–273.

Martel, A. T., and M. R. Gibling. 1996a. Stratigraphy and tectonic history of the Upper

Devonian to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic

Geology; Volume 32, Number 1 (1996)DO - 10.4138/2076 .

Martel, A. T., D. C. McGregor, and J. Utting. 1993. Stratigraphic significance of Upper

Devonian and Lower Carboniferous miospores from the type area of the Horton Group,

Nova Scotia. Canadian Journal of Earth Sciences 30:1091–1098.

Martel, T. A., and M. R. Gibling. 1996b. Stratigraphy and tectonic history of the Upper

Devonian to Lower Carboniferous Horton Bluff Formation, Nova Scotia. Atlantic Geology

32:13–38.

188

Meunier, F. J., and M. Laurin. 2012. A microanatomical and histological study of the fin long

bones of the Devonian sarcopterygian Eusthenopteron foordi. Acta Zoologica 93:88–97.

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2017. Reconstructing pectoral

appendicular muscle anatomy in fossil fish and tetrapods over the fins-to-limbs transition.

Biological Reviews. 93:1077-1107

Molnar, J. L., R. Diogo, J. R. Hutchinson, and S. E. Pierce. 2018. Evolution of Hindlimb Muscle

Anatomy Across the Tetrapod Water-to-Land Transition, Including Comparisons with

Forelimb Anatomy. Anatomical Record. doi: 10.1002/ar.23997.

Mossman, D. J., and R. G. Grantham. 2008. Eochelysipus horni, a new vertebrate trace fossil

from the Tournaisian Horton Bluff Formation, Nova Scotia. Atlantic Geology 44:69–77.

Nakajima, Y., and H. Endo. 2013. Comparative Humeral Microanatomy of Terrestrial,

Semiaquatic, and Aquatic Carnivorans using Micro-Focus CT Scan. Mammal Study 38:1–

8.

Neill, W. T. 1963. Notes on the Alabama Waterdog , Necturus alabamensis Viosca.

Herpetologica 19:166–174.

Otoo, B. K. A., J. A. Clack, T. R. Smithson, C. E. Bennett, T. I. Kearsey, and M. I. Coates. 2019.

A fish and tetrapod fauna from Romer’s Gap preserved in Scottish Tournaisian floodplain

deposits. Palaeontology 62:225–253.

Padian, K., and E.-T. Lamm. (Eds.). 2013. Bone Histology of Fossil Tetrapods: Advancing

Methods, Analysis, and Interpretation. University of California Press, Berkeley, California,

298 pp.

189

Panchen, A. L. 1985. On the Amphibian Crassigyrinus scoticus Watson from the Carboniferous

of Scotland. Philosophical Transactions of the Royal Society of London. Series B

309:505–568.

Panchen, A. L., and T. R. Smithson. 1990. The pelvic girdle and hind limb of Crassigyrinus

scoticus (Lydekker) from the Scottish Carboniferous and the origin of the tetrapod pelvic

skeleton. Transactions of the Royal Society of Edinburgh: Earth Sciences 81:31–44.

Petermann, H., and M. Sander. 2013. Histological evidence for muscle insertion in extant

amniote femora: Implications for muscle reconstruction in fossils. Journal of Anatomy

222:419–436.

Pierce, S. E., J. A. Clack, and J. R. Hutchinson. 2012. Three-dimensional limb joint mobility in

the early tetrapod Ichthyostega. Nature 486:523–526.

Plasse, M., E. Amson, J. Bardin, Q. Grimal, and D. Germain. 2019. Trabecular architecture in

the humeral metaphyses of non‐avian reptiles (Crocodylia, Squamata and Testudines):

Lifestyle, allometry and phylogeny. Journal of Morphology jmor.20996.

Quemeneur, S., V. de Buffrénil, and M. Laurin. 2013. Microanatomy of the amniote femur and

inference of lifestyle in limbed vertebrates. Biological Journal of the Linnean Society

109:644–655.

Romer, A. S. 1956. The Early Evolution of Land Vertebrates. Proceedings of the American

Philosophical Society 100:157–167.

190

Romer, A. S. 1957. The appendicular skeleton of the Permian embolomerous amphibian

Archeria. Contributions from the Museum of Paleontology, University of Michigan

13:103–159.

S. J. Godfrey. 1989. The Postcranial Skeletal Anatomy of the Carboniferous Tetrapod

Greererpeton burkemorani Romer, 1969. Philosophical Transactions of the Royal Society

B: Biological Sciences 323:75–133.

Ryan, T. M., and R. A. Ketcham. 2002a. The three-dimensional structure of trabecular bone in

the femoral head of strepsirrhine primates. Journal of Human Evolution 43:1–26.

Ryan, T. M., and R. A. Ketcham. 2002b. Femoral head trabecular bone structure in two

omomyid primates. Journal of Human Evolution 43:241–263.

Ryan, T. M., and R. A. Ketcham. 2005. Angular orientation of trabecular bone in the femoral

head and its relationship to hip joint loads in leaping primates. Journal of Morphology

265:249–263.

Sallan, L. C., M. I. Coates, J. Clack, L. C. Saltan, and M. I. Coatesa. 2010. End-Devonian

extinction and a bottleneck in the early evolution of modern jawed End-Devonian

extinction and a bottleneck in the early evolution of modern jawed vertebrates. Proceedings

of the National Academy of Sciences of the United States of America 107:10131–10135.

Sanchez, S., V. Dupret, P. Tafforeau, K. M. Trinajstic, B. Ryll, P. J. Gouttenoire, L. Wretman, L.

Zylberberg, F. Peyrin, and P. E. Ahlberg. 2013. 3D Microstructural Architecture of Muscle

Attachments in Extant and Fossil Vertebrates Revealed by Synchrotron Microtomography.

PLoS ONE 8:e56992.

191

Sanchez, S., P. Tafforeau, J. A. Clack, and P. E. Ahlberg. 2016. Life history of the stem tetrapod

Acanthostega revealed by synchrotron microtomography. Nature 537:408–411.

Sanchez, S., P. E. Ahlberg, K. M. Trinajstic, A. Mirone, and P. Tafforeau. 2012. Three-

dimensional synchrotron virtual paleohistology: A new insight into the world of fossil bone

microstructures. Microscopy and Microanalysis 18:1095–1105.

Sanchez, S., D. Germain, A. De Ricqlès, A. Abourachid, F. Goussard, and P. Tafforeau. 2010.

Limb-bone histology of temnospondyls: Implications for understanding the diversification

of palaeoecologies and patterns of locomotion of Permo-Triassic tetrapods. Journal of

Evolutionary Biology 23:2076–2090.

Shanbhag, A. G. 1994. Utilization of Information Measure as a Means of Image Thresholding.

CVGIP: Graphical Models and Image Processing 56:414–419.

Shubin, N. H., E. B. Daeschler, and F. A. Jenkins. 2014. Pelvic girdle and fin of Tiktaalik roseae.

Proceedings of the National Academy of Sciences of the United States of America

111:893–899.

Smithson, T. R. 1985. The morphology and relationships of the Carboniferous amphibian

Eoherpeton watsoni Panchen. Zoological Journal of the Linnean Society 85:317–410.

Smithson, T. R., and J. A. Clack. 2013. Tetrapod appendicular skeletal elements from the Early

Carboniferous of Scotland. Comptes Rendus - Palevol 12:405–417.

Smithson, T. R., and J. A. Clack. 2018. A new tetrapod from Romer’s Gap reveals an early

adaptation for walking. Earth and Environmental Science Transactions of the Royal

Society of Edinburgh 108:89–97.

192

Smithson, T. R., S. P. Wood, J. E. A. Marshall, and J. A. Clack. 2012. Earliest Carboniferous

tetrapod and arthropod faunas from Scotland populate Romer’s Gap. Proceedings of the

National Academy of Sciences of the United States of America 109:4532–7.

Sparreboom, M., and H. Ota. 1995. Notes on the life-history and reproductive behaviour of

Cynops ensicauda popei (Amphibia: Salamandridae). Herpetological Journal 5:310–315.

Su, A., and K. J. Carlson. 2017. Comparative analysis of trabecular bone structure and

orientation in South African hominin tali. Journal of Human Evolution 106:1–18.

Sues, H.-D., R. W. Hook, and P. E. Olsen. 2013. Donald Baird and his discoveries of

Carboniferous and early Mesozoic vertebrates in Nova Scotia. Atlantic Geology 49:90.

Tibert, N. E., and D. B. Scott. 1999. Ostracodes and Agglutinated Foraminifera as Indicators of

Paleoenvironmental Change in an Early Carboniferous Brackish Bay, Atlantic Canada.

PALAIOS 14:246.

Utting, J., J. D. Keppie, and P. S. Giles. 1989. Palynology and Stratigraphy of the Lower

Carboniferous Horton Group, Nova Scotia. Contributions to Canadian Paleontology,

Geological Survey of Canada Bulletin 3:117–143.

Wall, W. P. 1983. The correlation between high limb-bone density and aquatic habits in Recent

mammals (cetaceans, pinnipeds). Journal of Paleontology 57:197–207.

Walthall, J. C., and M. A. Ashley-Ross. 2006. Postcranial myology of the California newt,

Taricha torosa. 46–57.

193

Ward, P., C. Labandeira, M. Laurin, and R. A. Berner. 2006. Confirmation of Romer’s Gap as a

low oxygen interval constraining the timing of initial arthropod and vertebrate

terrestrialization. Proceedings of the National Academy of Sciences 103:16818–16822.

Warren, A. 2007. New data on Ossinodus pueri, a stem tetrapod from the Early Carboniferous of

Australia. Journal of Vertebrate Paleontology 27:850–862.

Warren, A., and S. Turner. 2004. The first stem tetrapod from the Lower Carboniferous of

Gondwana. Palaeontology 47:151–184.

Waters, C. N., R. A. Waters, W. J. Barclay, and J. R. Davies. 2009. A Lithostratigraphical

Framework for the Carboniferous Successions of Northern Great Britain (Onshore). 174

pp.

Williams-Hatala, E. M., K. G. Hatala, M. Gordon, A. Key, M. Kasper, and T. L. Kivell. 2018.

The manual pressures of stone tool behaviors and their implications for the evolution of the

human hand. Journal of Human Evolution 119:14–26.

Witmer, L. M. 1995. The Extant Phylogenetic Bracket and the importance of reconstructing soft

tissues in fossils; pp. 19–33 in Jeff Thomason (ed.), Functional morphology in vertebrate

paleontology. Cambridge University Press.

Zeininger, A., B. A. Patel, B. Zipfel, and K. J. Carlson. 2016. Trabecular architecture in the StW

352 fossil hominin calcaneus. Journal of Human Evolution 97:145–158.

194

Appendix A

Lateral view of Eublepharis (terrestrial) hindlimb musculature. Few of the muscles visibly interact directly with the bone, and tendons are not observed using diceCT methods.

195

Lateral view of Necturus (aquatic) hindlimb musculature. Few of the muscles visibly interact directly with the bone, and tendons are not observed using diceCT methods.

196

Modeled compactness at the periphery=0.971 AIC=720.8818 Appendix B R/t (Currey & Alexander 1985) : 3.3322 r2=0.5115007 CDI (Castanet et al. 2000) : 0.3001 Observed compactness=0.798 NSM.005.GF.045.044 From 18 ° to 23 ° Modeled compactness=0.795 Maximum likelihood:-Ln L=498.4562 Modeled compactness at the center=0.488

AIC=1004.912 Modeled compactness at the periphery=0.995 Global analysis r2=0.2456495 R/t (Currey & Alexander 1985) : 2.5798 Maximum likelihood:-Ln L=1223.987 Observed compactness=0.658 CDI (Castanet et al. 2000) : 0.3876 AIC=2455.973 Modeled compactness=0.658 From 48 ° to 53 ° r2=0.9330103 Modeled compactness at the center=0.533 Maximum likelihood:-Ln L=419.6125 Observed compactness=0.732 Modeled compactness at the periphery=0.961 AIC=847.225 Modeled compactness=0.732 R/t (Currey & Alexander 1985) : 6.2115 r2=0.4558604 Modeled compactness at the center=0.276 CDI (Castanet et al. 2000) : 0.161 Observed compactness=0.798 Modeled compactness at the periphery=0.985 From 24 ° to 29 ° Modeled compactness=0.799 R/t (Currey & Alexander 1985) : 2.3535 Maximum likelihood:-Ln L=532.548 Modeled compactness at the center=0.528 CDI (Castanet et al. 2000) : 0.4249 AIC=1073.096 Modeled compactness at the periphery=0.999

r2=0.191992 R/t (Currey & Alexander 1985) : 2.7906 Angular analysis Observed compactness=0.671 CDI (Castanet et al. 2000) : 0.3584 From 0 ° to 5 ° Modeled compactness=0.67 From 54 ° to 59 ° Maximum likelihood:-Ln L=336.8444 Modeled compactness at the center=0.533 Maximum likelihood:-Ln L=463.7137 AIC=681.6888 Modeled compactness at the periphery=0.906 AIC=935.4273 r2=0.5714005 R/t (Currey & Alexander 1985) : 4.7901 r2=0.5458898 Observed compactness=0.811 CDI (Castanet et al. 2000) : 0.2088 Observed compactness=0.839 Modeled compactness=0.801 From 30 ° to 35 ° Modeled compactness=0.84 Modeled compactness at the center=0.288 Maximum likelihood:-Ln L=347.7229 Modeled compactness at the center=0.468 Modeled compactness at the periphery=0.996 AIC=703.4458 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 1.985 r2=0.2190945 R/t (Currey & Alexander 1985) : 2.0556 CDI (Castanet et al. 2000) : 0.5038 Observed compactness=0.778 CDI (Castanet et al. 2000) : 0.4865 From 6 ° to 11 ° Modeled compactness=0.78 From 60 ° to 65 ° Maximum likelihood:-Ln L=270.0854 Modeled compactness at the center=0.678 Maximum likelihood:-Ln L=341.9554 AIC=548.1709 Modeled compactness at the periphery=1. AIC=691.9108 r2=0.5495696 R/t (Currey & Alexander 1985) : 5.6677 r2=0.7652581 Observed compactness=0.703 CDI (Castanet et al. 2000) : 0.1764 Observed compactness=0.857 Modeled compactness=0.701 From 36 ° to 41 ° Modeled compactness=0.86 Modeled compactness at the center=0.405 Maximum likelihood:-Ln L=345.912 Modeled compactness at the center=0.189 Modeled compactness at the periphery=0.987 AIC=699.824 Modeled compactness at the periphery=0.989 R/t (Currey & Alexander 1985) : 3.2465 r2=0.4603326 R/t (Currey & Alexander 1985) : 1.3373 CDI (Castanet et al. 2000) : 0.308 Observed compactness=0.807 CDI (Castanet et al. 2000) : 0.7478 From 12 ° to 17 ° Modeled compactness=0.805 From 66 ° to 71 ° Maximum likelihood:-Ln L=397.0572 Modeled compactness at the center=0.497 Maximum likelihood:-Ln L=283.5548 AIC=802.1144 Modeled compactness at the periphery=0.969 AIC=575.1096 r2=0.5394976 R/t (Currey & Alexander 1985) : 2.4166 r2=0.7770223 Observed compactness=0.623 CDI (Castanet et al. 2000) : 0.4138 Observed compactness=0.814 Modeled compactness=0.626 From 42 ° to 47 ° Modeled compactness=0.815 Modeled compactness at the center=0.277 Maximum likelihood:-Ln L=356.4409 Modeled compactness at the center=0.204

197

Modeled compactness at the periphery=0.971 AIC=812.5046 Modeled compactness at the periphery=0.984 R/t (Currey & Alexander 1985) : 1.4 r2=0.5400354 R/t (Currey & Alexander 1985) : 1.3081 CDI (Castanet et al. 2000) : 0.7143 Observed compactness=0.821 CDI (Castanet et al. 2000) : 0.7645 From 72 ° to 77 ° Modeled compactness=0.822 From 126 ° to 131 ° Maximum likelihood:-Ln L=327.4094 Modeled compactness at the center=0.173 Maximum likelihood:-Ln L=271.2646 AIC=662.8187 Modeled compactness at the periphery=0.978 AIC=550.5293 r2=0.6129339 R/t (Currey & Alexander 1985) : 1.4223 r2=0.6128501 Observed compactness=0.686 CDI (Castanet et al. 2000) : 0.7031 Observed compactness=0.91 Modeled compactness=0.685 From 102 ° to 107 ° Modeled compactness=0.909 Modeled compactness at the center=0.178 Maximum likelihood:-Ln L=515.0214 Modeled compactness at the center=0.269 Modeled compactness at the periphery=0.892 AIC=1038.043 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 1.7339 r2=0.4301875 R/t (Currey & Alexander 1985) : 1.1994 CDI (Castanet et al. 2000) : 0.5767 Observed compactness=0.785 CDI (Castanet et al. 2000) : 0.8338 From 78 ° to 83 ° Modeled compactness=0.786 From 132 ° to 137 ° Maximum likelihood:-Ln L=265.2249 Modeled compactness at the center=0.364 Maximum likelihood:-Ln L=338.7424 AIC=538.4498 Modeled compactness at the periphery=0.968 AIC=685.4849 r2=0.7551215 R/t (Currey & Alexander 1985) : 1.9827 r2=0.6001405 Observed compactness=0.747 CDI (Castanet et al. 2000) : 0.5044 Observed compactness=0.825 Modeled compactness=0.747 From 108 ° to 113 ° Modeled compactness=0.832 Modeled compactness at the center=0.057 Maximum likelihood:-Ln L=427.3818 Modeled compactness at the center=0.233 Modeled compactness at the periphery=0.977 AIC=862.7636 Modeled compactness at the periphery=0.977 R/t (Currey & Alexander 1985) : 1.7622 r2=0.3988653 R/t (Currey & Alexander 1985) : 1.5486 CDI (Castanet et al. 2000) : 0.5675 Observed compactness=0.813 CDI (Castanet et al. 2000) : 0.6457 From 84 ° to 89 ° Modeled compactness=0.815 From 138 ° to 143 ° Maximum likelihood:-Ln L=165.2368 Modeled compactness at the center=0.52 Maximum likelihood:-Ln L=277.6598 AIC=338.4737 Modeled compactness at the periphery=0.996 AIC=563.3196 r2=0.8992286 R/t (Currey & Alexander 1985) : 2.5028 r2=0.6997821 Observed compactness=0.76 CDI (Castanet et al. 2000) : 0.3996 Observed compactness=0.819 Modeled compactness=0.768 From 114 ° to 119 ° Modeled compactness=0.817 Modeled compactness at the center=0.037 Maximum likelihood:-Ln L=314.982 Modeled compactness at the center=0.289 Modeled compactness at the periphery=0.999 AIC=637.9639 Modeled compactness at the periphery=0.997 R/t (Currey & Alexander 1985) : 1.8751 r2=0.4900636 R/t (Currey & Alexander 1985) : 1.8985 CDI (Castanet et al. 2000) : 0.5333 Observed compactness=0.87 CDI (Castanet et al. 2000) : 0.5267 From 90 ° to 95 ° Modeled compactness=0.872 From 144 ° to 149 ° Maximum likelihood:-Ln L=397.2068 Modeled compactness at the center=0.436 Maximum likelihood:-Ln L=360.1568 AIC=802.4135 Modeled compactness at the periphery=0.997 AIC=728.3137 r2=0.4692395 R/t (Currey & Alexander 1985) : 1.7505 r2=0.7456744 Observed compactness=0.764 CDI (Castanet et al. 2000) : 0.5713 Observed compactness=0.756 Modeled compactness=0.765 From 120 ° to 125 ° Modeled compactness=0.755 Modeled compactness at the center=0.488 Maximum likelihood:-Ln L=396.8458 Modeled compactness at the center=0.101 Modeled compactness at the periphery=1. AIC=801.6915 Modeled compactness at the periphery=0.959 R/t (Currey & Alexander 1985) : 3.0164 r2=0.5765252 R/t (Currey & Alexander 1985) : 1.6447 CDI (Castanet et al. 2000) : 0.3315 Observed compactness=0.852 CDI (Castanet et al. 2000) : 0.608 From 96 ° to 101 ° Modeled compactness=0.855 From 150 ° to 155 ° Maximum likelihood:-Ln L=402.2523 Modeled compactness at the center=0.229 Maximum likelihood:-Ln L=480.003

198

AIC=968.006 Modeled compactness at the periphery=1. AIC=386.7069 r2=0.5708458 R/t (Currey & Alexander 1985) : 3.1257 r2=0.866489 Observed compactness=0.772 CDI (Castanet et al. 2000) : 0.3199 Observed compactness=0.636 Modeled compactness=0.776 From 180 ° to 185 ° Modeled compactness=0.643 Modeled compactness at the center=0.2 Maximum likelihood:-Ln L=214.0532 Modeled compactness at the center=0.001 Modeled compactness at the periphery=0.951 AIC=436.1063 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 1.4782 r2=0.8551919 R/t (Currey & Alexander 1985) : 2.3696 CDI (Castanet et al. 2000) : 0.6765 Observed compactness=0.698 CDI (Castanet et al. 2000) : 0.422 From 156 ° to 161 ° Modeled compactness=0.683 From 210 ° to 215 ° Maximum likelihood:-Ln L=445.6508 Modeled compactness at the center=0.047 Maximum likelihood:-Ln L=214.8333 AIC=899.3017 Modeled compactness at the periphery=0.997 AIC=437.6665 r2=0.5319054 R/t (Currey & Alexander 1985) : 2.2533 r2=0.840123 Observed compactness=0.685 CDI (Castanet et al. 2000) : 0.4438 Observed compactness=0.64 Modeled compactness=0.683 From 186 ° to 191 ° Modeled compactness=0.636 Modeled compactness at the center=0.325 Maximum likelihood:-Ln L=179.0319 Modeled compactness at the center=0.001 Modeled compactness at the periphery=0.99 AIC=366.0638 Modeled compactness at the periphery=0.992 R/t (Currey & Alexander 1985) : 3.022 r2=0.8751208 R/t (Currey & Alexander 1985) : 2.3881 CDI (Castanet et al. 2000) : 0.3309 Observed compactness=0.722 CDI (Castanet et al. 2000) : 0.4188 From 162 ° to 167 ° Modeled compactness=0.717 From 216 ° to 221 ° Maximum likelihood:-Ln L=271.3488 Modeled compactness at the center=0.001 Maximum likelihood:-Ln L=231.3629 AIC=550.6977 Modeled compactness at the periphery=0.999 AIC=470.7257 r2=0.6404043 R/t (Currey & Alexander 1985) : 2.0563 r2=0.833602 Observed compactness=0.687 CDI (Castanet et al. 2000) : 0.4863 Observed compactness=0.699 Modeled compactness=0.686 From 192 ° to 197 ° Modeled compactness=0.698 Modeled compactness at the center=0.364 Maximum likelihood:-Ln L=97.27199 Modeled compactness at the center=0.004 Modeled compactness at the periphery=1. AIC=202.544 Modeled compactness at the periphery=0.994 R/t (Currey & Alexander 1985) : 3.3183 r2=0.9371238 R/t (Currey & Alexander 1985) : 2.0887 CDI (Castanet et al. 2000) : 0.3014 Observed compactness=0.782 CDI (Castanet et al. 2000) : 0.4788 From 168 ° to 173 ° Modeled compactness=0.783 From 222 ° to 227 ° Maximum likelihood:-Ln L=251.0615 Modeled compactness at the center=0. Maximum likelihood:-Ln L=184.6184 AIC=510.123 Modeled compactness at the periphery=1. AIC=377.2367 r2=0.7107702 R/t (Currey & Alexander 1985) : 1.8386 r2=0.8833495 Observed compactness=0.743 CDI (Castanet et al. 2000) : 0.5439 Observed compactness=0.624 Modeled compactness=0.74 From 198 ° to 203 ° Modeled compactness=0.618 Modeled compactness at the center=0.277 Maximum likelihood:-Ln L=152.8417 Modeled compactness at the center=0. Modeled compactness at the periphery=0.997 AIC=313.6834 Modeled compactness at the periphery=0.999 R/t (Currey & Alexander 1985) : 2.3885 r2=0.9073991 R/t (Currey & Alexander 1985) : 2.5315 CDI (Castanet et al. 2000) : 0.4187 Observed compactness=0.714 CDI (Castanet et al. 2000) : 0.395 From 174 ° to 179 ° Modeled compactness=0.724 From 228 ° to 233 ° Maximum likelihood:-Ln L=253.4359 Modeled compactness at the center=0.001 Maximum likelihood:-Ln L=225.9586 AIC=514.8718 Modeled compactness at the periphery=0.998 AIC=459.9172 r2=0.5834724 R/t (Currey & Alexander 1985) : 2.0171 r2=0.8486022 Observed compactness=0.692 CDI (Castanet et al. 2000) : 0.4958 Observed compactness=0.599 Modeled compactness=0.713 From 204 ° to 209 ° Modeled compactness=0.592 Modeled compactness at the center=0.392 Maximum likelihood:-Ln L=189.3534 Modeled compactness at the center=0.

199

Modeled compactness at the periphery=0.936 AIC=443.0448 Modeled compactness at the periphery=0.999 R/t (Currey & Alexander 1985) : 2.4437 r2=0.8703535 R/t (Currey & Alexander 1985) : 2.3173 CDI (Castanet et al. 2000) : 0.4092 Observed compactness=0.671 CDI (Castanet et al. 2000) : 0.4315 From 234 ° to 239 ° Modeled compactness=0.667 From 288 ° to 293 ° Maximum likelihood:-Ln L=319.8372 Modeled compactness at the center=0.004 Maximum likelihood:-Ln L=159.4437 AIC=647.6744 Modeled compactness at the periphery=0.989 AIC=326.8874 r2=0.7615629 R/t (Currey & Alexander 1985) : 2.2085 r2=0.8784989 Observed compactness=0.627 CDI (Castanet et al. 2000) : 0.4528 Observed compactness=0.666 Modeled compactness=0.618 From 264 ° to 269 ° Modeled compactness=0.67 Modeled compactness at the center=0.007 Maximum likelihood:-Ln L=251.0692 Modeled compactness at the center=0.025 Modeled compactness at the periphery=0.971 AIC=510.1385 Modeled compactness at the periphery=0.999 R/t (Currey & Alexander 1985) : 2.4053 r2=0.845093 R/t (Currey & Alexander 1985) : 2.3076 CDI (Castanet et al. 2000) : 0.4157 Observed compactness=0.595 CDI (Castanet et al. 2000) : 0.4334 From 240 ° to 245 ° Modeled compactness=0.6 From 294 ° to 299 ° Maximum likelihood:-Ln L=162.4278 Modeled compactness at the center=0.042 Maximum likelihood:-Ln L=148.2358 AIC=332.8556 Modeled compactness at the periphery=0.987 AIC=304.4717 r2=0.9220979 R/t (Currey & Alexander 1985) : 2.6734 r2=0.8578429 Observed compactness=0.617 CDI (Castanet et al. 2000) : 0.3741 Observed compactness=0.66 Modeled compactness=0.617 From 270 ° to 275 ° Modeled compactness=0.655 Modeled compactness at the center=0.001 Maximum likelihood:-Ln L=214.6362 Modeled compactness at the center=0.005 Modeled compactness at the periphery=0.991 AIC=437.2724 Modeled compactness at the periphery=0.986 R/t (Currey & Alexander 1985) : 2.485 r2=0.8743339 R/t (Currey & Alexander 1985) : 2.2556 CDI (Castanet et al. 2000) : 0.4024 Observed compactness=0.642 CDI (Castanet et al. 2000) : 0.4433 From 246 ° to 251 ° Modeled compactness=0.648 From 300 ° to 305 ° Maximum likelihood:-Ln L=334.1069 Modeled compactness at the center=0.029 Maximum likelihood:-Ln L=160.1123 AIC=676.2138 Modeled compactness at the periphery=0.947 AIC=328.2245 r2=0.7600563 R/t (Currey & Alexander 1985) : 2.2668 r2=0.8542445 Observed compactness=0.612 CDI (Castanet et al. 2000) : 0.4411 Observed compactness=0.694 Modeled compactness=0.612 From 276 ° to 281 ° Modeled compactness=0.691 Modeled compactness at the center=0.007 Maximum likelihood:-Ln L=85.81169 Modeled compactness at the center=0.004 Modeled compactness at the periphery=0.971 AIC=179.6234 Modeled compactness at the periphery=0.993 R/t (Currey & Alexander 1985) : 2.439 r2=0.9259182 R/t (Currey & Alexander 1985) : 2.1176 CDI (Castanet et al. 2000) : 0.41 Observed compactness=0.68 CDI (Castanet et al. 2000) : 0.4722 From 252 ° to 257 ° Modeled compactness=0.676 From 306 ° to 311 ° Maximum likelihood:-Ln L=256.1705 Modeled compactness at the center=0.011 Maximum likelihood:-Ln L=286.7428 AIC=520.341 Modeled compactness at the periphery=1. AIC=581.4857 r2=0.8491449 R/t (Currey & Alexander 1985) : 2.2836 r2=0.6555576 Observed compactness=0.634 CDI (Castanet et al. 2000) : 0.4379 Observed compactness=0.581 Modeled compactness=0.631 From 282 ° to 287 ° Modeled compactness=0.573 Modeled compactness at the center=0.002 Maximum likelihood:-Ln L=156.633 Modeled compactness at the center=0.053 Modeled compactness at the periphery=0.987 AIC=321.2661 Modeled compactness at the periphery=0.96 R/t (Currey & Alexander 1985) : 2.391 r2=0.8715613 R/t (Currey & Alexander 1985) : 2.8074 CDI (Castanet et al. 2000) : 0.4182 Observed compactness=0.663 CDI (Castanet et al. 2000) : 0.3562 From 258 ° to 263 ° Modeled compactness=0.662 From 312 ° to 317 ° Maximum likelihood:-Ln L=217.5224 Modeled compactness at the center=0. Maximum likelihood:-Ln L=244.2674

200

AIC=496.5348 Modeled compactness at the periphery=0.964 Modeled compactness at the periphery=0.983 r2=0.7299128 R/t (Currey & Alexander 1985) : 3.2077 R/t (Currey & Alexander 1985) : 3.5338 Observed compactness=0.616 CDI (Castanet et al. 2000) : 0.3118 CDI (Castanet et al. 2000) : 0.283 Modeled compactness=0.591 From 342 ° to 347 ° Angular analysis Modeled compactness at the center=0.01 Maximum likelihood:-Ln L=289.4094 From 0 ° to 5 ° Modeled compactness at the periphery=0.974 AIC=586.8188 Maximum likelihood:-Ln L=664.1675 R/t (Currey & Alexander 1985) : 2.5953 r2=0.5555505 AIC=1336.335 CDI (Castanet et al. 2000) : 0.3853 Observed compactness=0.679 r2=0.1311276 From 318 ° to 323 ° Modeled compactness=0.672 Observed compactness=0.734 Maximum likelihood:-Ln L=221.5055 Modeled compactness at the center=0.069 Modeled compactness=0.731 AIC=451.011 Modeled compactness at the periphery=0.936 Modeled compactness at the center=0.637 r2=0.7757267 R/t (Currey & Alexander 1985) : 1.9796 Modeled compactness at the periphery=1. Observed compactness=0.553 CDI (Castanet et al. 2000) : 0.5052 R/t (Currey & Alexander 1985) : 6.8315 Modeled compactness=0.545 From 348 ° to 353 ° CDI (Castanet et al. 2000) : 0.1464 Modeled compactness at the center=0.002 Maximum likelihood:-Ln L=230.9777 From 6 ° to 11 ° Modeled compactness at the periphery=0.969 AIC=469.9554 Maximum likelihood:-Ln L=477.6519 R/t (Currey & Alexander 1985) : 2.8713 r2=0.7766837 AIC=963.3038 CDI (Castanet et al. 2000) : 0.3483 Observed compactness=0.551 r2=0.1966577 From 324 ° to 329 ° Modeled compactness=0.546 Observed compactness=0.748 Maximum likelihood:-Ln L=242.8205 Modeled compactness at the center=0.106 Modeled compactness=0.747 AIC=493.641 Modeled compactness at the periphery=0.959 Modeled compactness at the center=0.644 r2=0.679929 R/t (Currey & Alexander 1985) : 3.1876 Modeled compactness at the periphery=1. Observed compactness=0.575 CDI (Castanet et al. 2000) : 0.3137 R/t (Currey & Alexander 1985) : 6.2216 Modeled compactness=0.569 From 354 ° to 359 ° CDI (Castanet et al. 2000) : 0.1607 Modeled compactness at the center=0.016 Maximum likelihood:-Ln L=258.2134 From 12 ° to 17 ° Modeled compactness at the periphery=0.928 AIC=524.4268 Maximum likelihood:-Ln L=473.1652 R/t (Currey & Alexander 1985) : 2.6147 r2=0.7095752 AIC=954.3304 CDI (Castanet et al. 2000) : 0.3825 Observed compactness=0.594 r2=0.1903071 From 330 ° to 335 ° Modeled compactness=0.58 Observed compactness=0.712 Maximum likelihood:-Ln L=254.8574 Modeled compactness at the center=0.151 Modeled compactness=0.711 AIC=517.7147 Modeled compactness at the periphery=0.998 Modeled compactness at the center=0.602 r2=0.5818154 R/t (Currey & Alexander 1985) : 3.2777 Modeled compactness at the periphery=1. Observed compactness=0.539 CDI (Castanet et al. 2000) : 0.3051 R/t (Currey & Alexander 1985) : 6.6665 Modeled compactness=0.533 CDI (Castanet et al. 2000) : 0.15 Modeled compactness at the center=0.173 From 18 ° to 23 ° Modeled compactness at the periphery=0.943 NSM.005.GF.045.o47-D Maximum likelihood:-Ln L=545.5323 R/t (Currey & Alexander 1985) : 3.8016 AIC=1099.065 CDI (Castanet et al. 2000) : 0.263 Global analysis r2=0.2379584 From 336 ° to 341 ° Maximum likelihood:-Ln L=1719.766 Observed compactness=0.686 Maximum likelihood:-Ln L=339.2423 AIC=3447.532 Modeled compactness=0.685 AIC=686.4847 r2=0.864652 Modeled compactness at the center=0.563 r2=0.4781864 Observed compactness=0.706 Modeled compactness at the periphery=1. Observed compactness=0.643 Modeled compactness=0.706 R/t (Currey & Alexander 1985) : 6.61 Modeled compactness=0.643 Modeled compactness at the center=0.449 CDI (Castanet et al. 2000) : 0.1513 Modeled compactness at the center=0.305 From 24 ° to 29 °

201

Maximum likelihood:-Ln L=620.3998 Modeled compactness at the center=0.654 Maximum likelihood:-Ln L=371.5283 AIC=1248.8 Modeled compactness at the periphery=1. AIC=751.0566 r2=0.2045376 R/t (Currey & Alexander 1985) : 5.2381 r2=0.5135626 Observed compactness=0.692 CDI (Castanet et al. 2000) : 0.1909 Observed compactness=0.786 Modeled compactness=0.692 From 54 ° to 59 ° Modeled compactness=0.783 Modeled compactness at the center=0.572 Maximum likelihood:-Ln L=464.6731 Modeled compactness at the center=0.453 Modeled compactness at the periphery=0.998 AIC=937.3462 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 6.4612 r2=0.4183411 R/t (Currey & Alexander 1985) : 2.676 CDI (Castanet et al. 2000) : 0.1548 Observed compactness=0.814 CDI (Castanet et al. 2000) : 0.3737 From 30 ° to 35 ° Modeled compactness=0.816 From 84 ° to 89 ° Maximum likelihood:-Ln L=676.9926 Modeled compactness at the center=0.388 Maximum likelihood:-Ln L=317.8951 AIC=1361.985 Modeled compactness at the periphery=0.969 AIC=643.7901 r2=0.2703922 R/t (Currey & Alexander 1985) : 1.7431 r2=0.5683351 Observed compactness=0.681 CDI (Castanet et al. 2000) : 0.5737 Observed compactness=0.747 Modeled compactness=0.679 From 60 ° to 65 ° Modeled compactness=0.753 Modeled compactness at the center=0.528 Maximum likelihood:-Ln L=252.1024 Modeled compactness at the center=0.212 Modeled compactness at the periphery=1. AIC=512.2048 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 5.7379 r2=0.7397204 R/t (Currey & Alexander 1985) : 2.2181 CDI (Castanet et al. 2000) : 0.1743 Observed compactness=0.854 CDI (Castanet et al. 2000) : 0.4508 From 36 ° to 41 ° Modeled compactness=0.857 From 90 ° to 95 ° Maximum likelihood:-Ln L=589.8185 Modeled compactness at the center=0.227 Maximum likelihood:-Ln L=475.7453 AIC=1187.637 Modeled compactness at the periphery=0.969 AIC=959.4907 r2=0.2449852 R/t (Currey & Alexander 1985) : 1.6329 r2=0.1957932 Observed compactness=0.722 CDI (Castanet et al. 2000) : 0.6124 Observed compactness=0.755 Modeled compactness=0.719 From 66 ° to 71 ° Modeled compactness=0.754 Modeled compactness at the center=0.603 Maximum likelihood:-Ln L=232.5932 Modeled compactness at the center=0.656 Modeled compactness at the periphery=1. AIC=473.1864 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 6.0839 r2=0.8025321 R/t (Currey & Alexander 1985) : 6.3898 CDI (Castanet et al. 2000) : 0.1644 Observed compactness=0.778 CDI (Castanet et al. 2000) : 0.1565 From 42 ° to 47 ° Modeled compactness=0.782 From 96 ° to 101 ° Maximum likelihood:-Ln L=434.0611 Modeled compactness at the center=0.146 Maximum likelihood:-Ln L=666.0005 AIC=876.1221 Modeled compactness at the periphery=0.999 AIC=1340.001 r2=0.3299409 R/t (Currey & Alexander 1985) : 1.9435 r2=0.2462659 Observed compactness=0.754 CDI (Castanet et al. 2000) : 0.5145 Observed compactness=0.755 Modeled compactness=0.751 From 72 ° to 77 ° Modeled compactness=0.758 Modeled compactness at the center=0.541 Maximum likelihood:-Ln L=341.9117 Modeled compactness at the center=0.583 Modeled compactness at the periphery=1. AIC=691.8234 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 3.7286 r2=0.6540613 R/t (Currey & Alexander 1985) : 4.1338 CDI (Castanet et al. 2000) : 0.2682 Observed compactness=0.786 CDI (Castanet et al. 2000) : 0.2419 From 48 ° to 53 ° Modeled compactness=0.791 From 102 ° to 107 ° Maximum likelihood:-Ln L=559.2338 Modeled compactness at the center=0.223 Maximum likelihood:-Ln L=962.1953 AIC=1126.468 Modeled compactness at the periphery=1. AIC=1932.391 r2=0.2353655 R/t (Currey & Alexander 1985) : 2.0281 r2=0.1169776 Observed compactness=0.773 CDI (Castanet et al. 2000) : 0.4931 Observed compactness=0.76 Modeled compactness=0.772 From 78 ° to 83 ° Modeled compactness=0.76

202

Modeled compactness at the center=0.657 Maximum likelihood:-Ln L=386.8141 Modeled compactness at the center=0.125 Modeled compactness at the periphery=0.945 AIC=781.6282 Modeled compactness at the periphery=0.959 R/t (Currey & Alexander 1985) : 6.1455 r2=0.3404723 R/t (Currey & Alexander 1985) : 2.0778 CDI (Castanet et al. 2000) : 0.1627 Observed compactness=0.807 CDI (Castanet et al. 2000) : 0.4813 From 108 ° to 113 ° Modeled compactness=0.813 From 162 ° to 167 ° Maximum likelihood:-Ln L=573.3158 Modeled compactness at the center=0.418 Maximum likelihood:-Ln L=198.359 AIC=1154.632 Modeled compactness at the periphery=0.995 AIC=404.718 r2=0.5432822 R/t (Currey & Alexander 1985) : 2.1529 r2=0.5099216 Observed compactness=0.779 CDI (Castanet et al. 2000) : 0.4645 Observed compactness=0.685 Modeled compactness=0.779 From 138 ° to 143 ° Modeled compactness=0.69 Modeled compactness at the center=0.05 Maximum likelihood:-Ln L=197.1165 Modeled compactness at the center=0.4 Modeled compactness at the periphery=0.801 AIC=402.233 Modeled compactness at the periphery=0.973 R/t (Currey & Alexander 1985) : 1.1643 r2=0.4175894 R/t (Currey & Alexander 1985) : 3.3346 CDI (Castanet et al. 2000) : 0.8589 Observed compactness=0.691 CDI (Castanet et al. 2000) : 0.2999 From 114 ° to 119 ° Modeled compactness=0.693 From 168 ° to 173 ° Maximum likelihood:-Ln L=664.2664 Modeled compactness at the center=0.5 Maximum likelihood:-Ln L=302.1366 AIC=1336.533 Modeled compactness at the periphery=0.999 AIC=612.2731 r2=0.1184215 R/t (Currey & Alexander 1985) : 4.5265 r2=0.2851838 Observed compactness=0.718 CDI (Castanet et al. 2000) : 0.2209 Observed compactness=0.721 Modeled compactness=0.719 From 144 ° to 149 ° Modeled compactness=0.721 Modeled compactness at the center=0.672 Maximum likelihood:-Ln L=252.7793 Modeled compactness at the center=0.458 Modeled compactness at the periphery=0.947 AIC=513.5586 Modeled compactness at the periphery=0.831 R/t (Currey & Alexander 1985) : 11.0644 r2=0.3455693 R/t (Currey & Alexander 1985) : 2.2037 CDI (Castanet et al. 2000) : 0.0904 Observed compactness=0.709 CDI (Castanet et al. 2000) : 0.4538 From 120 ° to 125 ° Modeled compactness=0.712 From 174 ° to 179 ° Maximum likelihood:-Ln L=530.5469 Modeled compactness at the center=0.424 Maximum likelihood:-Ln L=331.3528 AIC=1069.094 Modeled compactness at the periphery=0.926 AIC=670.7056 r2=0.1962253 R/t (Currey & Alexander 1985) : 2.8017 r2=0.3385527 Observed compactness=0.724 CDI (Castanet et al. 2000) : 0.3569 Observed compactness=0.649 Modeled compactness=0.723 From 150 ° to 155 ° Modeled compactness=0.64 Modeled compactness at the center=0.658 Maximum likelihood:-Ln L=301.8085 Modeled compactness at the center=0.302 Modeled compactness at the periphery=1. AIC=611.617 Modeled compactness at the periphery=0.916 R/t (Currey & Alexander 1985) : 9.5074 r2=0.2412545 R/t (Currey & Alexander 1985) : 3.2101 CDI (Castanet et al. 2000) : 0.1052 Observed compactness=0.659 CDI (Castanet et al. 2000) : 0.3115 From 126 ° to 131 ° Modeled compactness=0.662 From 180 ° to 185 ° Maximum likelihood:-Ln L=471.2163 Modeled compactness at the center=0.552 Maximum likelihood:-Ln L=260.8282 AIC=950.4326 Modeled compactness at the periphery=1. AIC=529.6563 r2=0.2890947 R/t (Currey & Alexander 1985) : 7.6069 r2=0.8800214 Observed compactness=0.785 CDI (Castanet et al. 2000) : 0.1315 Observed compactness=0.575 Modeled compactness=0.784 From 156 ° to 161 ° Modeled compactness=0.564 Modeled compactness at the center=0.622 Maximum likelihood:-Ln L=291.9301 Modeled compactness at the center=0.102 Modeled compactness at the periphery=1. AIC=591.8602 Modeled compactness at the periphery=0.997 R/t (Currey & Alexander 1985) : 4.0407 r2=0.5400023 R/t (Currey & Alexander 1985) : 3.2024 CDI (Castanet et al. 2000) : 0.2475 Observed compactness=0.692 CDI (Castanet et al. 2000) : 0.3123 From 132 ° to 137 ° Modeled compactness=0.697 From 186 ° to 191 °

203

Maximum likelihood:-Ln L=391.1117 Modeled compactness at the center=0.274 Maximum likelihood:-Ln L=383.9103 AIC=790.2234 Modeled compactness at the periphery=0.99 AIC=775.8207 r2=0.7103117 R/t (Currey & Alexander 1985) : 2.6537 r2=0.4816322 Observed compactness=0.634 CDI (Castanet et al. 2000) : 0.3768 Observed compactness=0.741 Modeled compactness=0.641 From 216 ° to 221 ° Modeled compactness=0.738 Modeled compactness at the center=0.221 Maximum likelihood:-Ln L=530.3507 Modeled compactness at the center=0.421 Modeled compactness at the periphery=0.992 AIC=1068.701 Modeled compactness at the periphery=0.988 R/t (Currey & Alexander 1985) : 2.9821 r2=0.4920762 R/t (Currey & Alexander 1985) : 2.8916 CDI (Castanet et al. 2000) : 0.3353 Observed compactness=0.698 CDI (Castanet et al. 2000) : 0.3458 From 192 ° to 197 ° Modeled compactness=0.694 From 246 ° to 251 ° Maximum likelihood:-Ln L=456.4548 Modeled compactness at the center=0.362 Maximum likelihood:-Ln L=418.6146 AIC=920.9095 Modeled compactness at the periphery=1. AIC=845.2292 r2=0.7451516 R/t (Currey & Alexander 1985) : 3.184 r2=0.3724624 Observed compactness=0.593 CDI (Castanet et al. 2000) : 0.3141 Observed compactness=0.635 Modeled compactness=0.591 From 222 ° to 227 ° Modeled compactness=0.635 Modeled compactness at the center=0.117 Maximum likelihood:-Ln L=299.297 Modeled compactness at the center=0.435 Modeled compactness at the periphery=0.995 AIC=606.594 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 3.0199 r2=0.5671578 R/t (Currey & Alexander 1985) : 5.1096 CDI (Castanet et al. 2000) : 0.3311 Observed compactness=0.703 CDI (Castanet et al. 2000) : 0.1957 From 198 ° to 203 ° Modeled compactness=0.7 From 252 ° to 257 ° Maximum likelihood:-Ln L=394.6922 Modeled compactness at the center=0.437 Maximum likelihood:-Ln L=330.4869 AIC=797.3845 Modeled compactness at the periphery=1. AIC=668.9738 r2=0.7318643 R/t (Currey & Alexander 1985) : 3.6403 r2=0.3867846 Observed compactness=0.652 CDI (Castanet et al. 2000) : 0.2747 Observed compactness=0.66 Modeled compactness=0.653 From 228 ° to 233 ° Modeled compactness=0.662 Modeled compactness at the center=0.198 Maximum likelihood:-Ln L=294.4677 Modeled compactness at the center=0.475 Modeled compactness at the periphery=0.998 AIC=596.9354 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 2.8356 r2=0.4680551 R/t (Currey & Alexander 1985) : 4.9928 CDI (Castanet et al. 2000) : 0.3527 Observed compactness=0.723 CDI (Castanet et al. 2000) : 0.2003 From 204 ° to 209 ° Modeled compactness=0.719 From 258 ° to 263 ° Maximum likelihood:-Ln L=473.6786 Modeled compactness at the center=0.466 Maximum likelihood:-Ln L=339.678 AIC=955.3572 Modeled compactness at the periphery=1. AIC=687.356 r2=0.6196021 R/t (Currey & Alexander 1985) : 3.5775 r2=0.5630058 Observed compactness=0.686 CDI (Castanet et al. 2000) : 0.2795 Observed compactness=0.751 Modeled compactness=0.697 From 234 ° to 239 ° Modeled compactness=0.748 Modeled compactness at the center=0.332 Maximum likelihood:-Ln L=302.1642 Modeled compactness at the center=0.155 Modeled compactness at the periphery=1. AIC=612.3283 Modeled compactness at the periphery=0.946 R/t (Currey & Alexander 1985) : 3.011 r2=0.7077423 R/t (Currey & Alexander 1985) : 1.6015 CDI (Castanet et al. 2000) : 0.3321 Observed compactness=0.73 CDI (Castanet et al. 2000) : 0.6244 From 210 ° to 215 ° Modeled compactness=0.727 From 264 ° to 269 ° Maximum likelihood:-Ln L=483.3135 Modeled compactness at the center=0.056 Maximum likelihood:-Ln L=288.9056 AIC=974.6271 Modeled compactness at the periphery=0.97 AIC=585.8112 r2=0.6230929 R/t (Currey & Alexander 1985) : 1.8226 r2=0.5294385 Observed compactness=0.701 CDI (Castanet et al. 2000) : 0.5487 Observed compactness=0.561 Modeled compactness=0.7 From 240 ° to 245 ° Modeled compactness=0.562

204

Modeled compactness at the center=0.286 Maximum likelihood:-Ln L=385.1968 Modeled compactness at the center=0.485 Modeled compactness at the periphery=0.988 AIC=778.3936 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 4.4573 r2=0.6640039 R/t (Currey & Alexander 1985) : 4.7656 CDI (Castanet et al. 2000) : 0.2244 Observed compactness=0.613 CDI (Castanet et al. 2000) : 0.2098 From 270 ° to 275 ° Modeled compactness=0.614 From 324 ° to 329 ° Maximum likelihood:-Ln L=488.3914 Modeled compactness at the center=0.223 Maximum likelihood:-Ln L=387.9452 AIC=984.7829 Modeled compactness at the periphery=1. AIC=783.8904 r2=0.4625986 R/t (Currey & Alexander 1985) : 3.3183 r2=0.3525052 Observed compactness=0.589 CDI (Castanet et al. 2000) : 0.3014 Observed compactness=0.67 Modeled compactness=0.59 From 300 ° to 305 ° Modeled compactness=0.668 Modeled compactness at the center=0.392 Maximum likelihood:-Ln L=440.7639 Modeled compactness at the center=0.507 Modeled compactness at the periphery=1. AIC=889.5278 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 5.4873 r2=0.5008729 R/t (Currey & Alexander 1985) : 5.467 CDI (Castanet et al. 2000) : 0.1822 Observed compactness=0.631 CDI (Castanet et al. 2000) : 0.1829 From 276 ° to 281 ° Modeled compactness=0.627 From 330 ° to 335 ° Maximum likelihood:-Ln L=413.2383 Modeled compactness at the center=0.271 Maximum likelihood:-Ln L=477.619 AIC=834.4766 Modeled compactness at the periphery=0.997 AIC=963.238 r2=0.5626382 R/t (Currey & Alexander 1985) : 3.4066 r2=0.4645208 Observed compactness=0.578 CDI (Castanet et al. 2000) : 0.2935 Observed compactness=0.695 Modeled compactness=0.577 From 306 ° to 311 ° Modeled compactness=0.699 Modeled compactness at the center=0.302 Maximum likelihood:-Ln L=280.1647 Modeled compactness at the center=0.277 Modeled compactness at the periphery=0.999 AIC=568.3293 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 4.4227 r2=0.6506465 R/t (Currey & Alexander 1985) : 2.6987 CDI (Castanet et al. 2000) : 0.2261 Observed compactness=0.678 CDI (Castanet et al. 2000) : 0.3706 From 282 ° to 287 ° Modeled compactness=0.671 From 336 ° to 341 ° Maximum likelihood:-Ln L=369.1576 Modeled compactness at the center=0.271 Maximum likelihood:-Ln L=425.5521 AIC=746.3151 Modeled compactness at the periphery=1. AIC=859.1042 r2=0.5741636 R/t (Currey & Alexander 1985) : 2.9753 r2=0.4521046 Observed compactness=0.569 CDI (Castanet et al. 2000) : 0.3361 Observed compactness=0.646 Modeled compactness=0.572 From 312 ° to 317 ° Modeled compactness=0.653 Modeled compactness at the center=0.336 Maximum likelihood:-Ln L=266.7268 Modeled compactness at the center=0.322 Modeled compactness at the periphery=1. AIC=541.4536 Modeled compactness at the periphery=0.997 R/t (Currey & Alexander 1985) : 5.0614 r2=0.7437451 R/t (Currey & Alexander 1985) : 3.4066 CDI (Castanet et al. 2000) : 0.1976 Observed compactness=0.686 CDI (Castanet et al. 2000) : 0.2935 From 288 ° to 293 ° Modeled compactness=0.669 From 342 ° to 347 ° Maximum likelihood:-Ln L=323.3104 Modeled compactness at the center=0.075 Maximum likelihood:-Ln L=404.3738 AIC=654.6208 Modeled compactness at the periphery=0.996 AIC=816.7477 r2=0.7335353 R/t (Currey & Alexander 1985) : 2.3818 r2=0.4244532 Observed compactness=0.611 CDI (Castanet et al. 2000) : 0.4199 Observed compactness=0.68 Modeled compactness=0.616 From 318 ° to 323 ° Modeled compactness=0.679 Modeled compactness at the center=0.212 Maximum likelihood:-Ln L=340.2386 Modeled compactness at the center=0.302 Modeled compactness at the periphery=1. AIC=688.4772 Modeled compactness at the periphery=0.997 R/t (Currey & Alexander 1985) : 3.2814 r2=0.4177969 R/t (Currey & Alexander 1985) : 2.9994 CDI (Castanet et al. 2000) : 0.3047 Observed compactness=0.681 CDI (Castanet et al. 2000) : 0.3334 From 294 ° to 299 ° Modeled compactness=0.676 From 348 ° to 353 °

205

Maximum likelihood:-Ln L=434.3855 CDI (Castanet et al. 2000) : 0.4222 Observed compactness=0.697 AIC=876.7711 From 6 ° to 11 ° Modeled compactness=0.694 r2=0.3161018 Maximum likelihood:-Ln L=271.5182 Modeled compactness at the center=0.329 Observed compactness=0.698 AIC=551.0365 Modeled compactness at the periphery=0.962 Modeled compactness=0.697 r2=0.6011457 R/t (Currey & Alexander 1985) : 2.7896 Modeled compactness at the center=0.53 Observed compactness=0.648 CDI (Castanet et al. 2000) : 0.3585 Modeled compactness at the periphery=1. Modeled compactness=0.65 From 36 ° to 41 ° R/t (Currey & Alexander 1985) : 4.9928 Modeled compactness at the center=0.221 Maximum likelihood:-Ln L=435.3696 CDI (Castanet et al. 2000) : 0.2003 Modeled compactness at the periphery=0.974 AIC=878.7391 From 354 ° to 359 ° R/t (Currey & Alexander 1985) : 2.8107 r2=0.3127028 Maximum likelihood:-Ln L=343.9329 CDI (Castanet et al. 2000) : 0.3558 Observed compactness=0.681 AIC=695.8658 From 12 ° to 17 ° Modeled compactness=0.677 r2=0.4938216 Maximum likelihood:-Ln L=343.9464 Modeled compactness at the center=0.339 Observed compactness=0.716 AIC=695.8928 Modeled compactness at the periphery=0.928 Modeled compactness=0.707 r2=0.6501501 R/t (Currey & Alexander 1985) : 2.9447 Modeled compactness at the center=0.306 Observed compactness=0.691 CDI (Castanet et al. 2000) : 0.3396 Modeled compactness at the periphery=0.997 Modeled compactness=0.687 From 42 ° to 47 ° R/t (Currey & Alexander 1985) : 2.7449 Modeled compactness at the center=0.145 Maximum likelihood:-Ln L=351.0375 CDI (Castanet et al. 2000) : 0.3643 Modeled compactness at the periphery=0.978 AIC=710.0751 R/t (Currey & Alexander 1985) : 2.2991 r2=0.3316982 CDI (Castanet et al. 2000) : 0.435 Observed compactness=0.739 NSM.007.GF.004.630 From 18 ° to 23 ° Modeled compactness=0.736 Maximum likelihood:-Ln L=379.7088 Modeled compactness at the center=0.317 Global analysis AIC=767.4177 Modeled compactness at the periphery=0.918 Maximum likelihood:-Ln L=1884.098 r2=0.4720502 R/t (Currey & Alexander 1985) : 1.8578 AIC=3776.195 Observed compactness=0.636 CDI (Castanet et al. 2000) : 0.5383 r2=0.8521959 Modeled compactness=0.633 From 48 ° to 53 ° Observed compactness=0.736 Modeled compactness at the center=0.374 Maximum likelihood:-Ln L=312.7015 Modeled compactness=0.736 Modeled compactness at the periphery=0.986 AIC=633.403 Modeled compactness at the center=0.4 R/t (Currey & Alexander 1985) : 4.0656 r2=0.5517284 Modeled compactness at the periphery=0.933 CDI (Castanet et al. 2000) : 0.246 Observed compactness=0.737 R/t (Currey & Alexander 1985) : 2.4494 From 24 ° to 29 ° Modeled compactness=0.733 CDI (Castanet et al. 2000) : 0.4083 Maximum likelihood:-Ln L=347.5424 Modeled compactness at the center=0.159 AIC=703.0847 Modeled compactness at the periphery=0.971 Angular analysis r2=0.557508 R/t (Currey & Alexander 1985) : 1.9618 From 0 ° to 5 ° Observed compactness=0.699 CDI (Castanet et al. 2000) : 0.5097 Maximum likelihood:-Ln L=290.6414 Modeled compactness=0.707 From 54 ° to 59 ° AIC=589.2828 Modeled compactness at the center=0.252 Maximum likelihood:-Ln L=334.6572 r2=0.5889634 Modeled compactness at the periphery=0.964 AIC=677.3144 Observed compactness=0.682 R/t (Currey & Alexander 1985) : 2.3573 r2=0.5862562 Modeled compactness=0.71 CDI (Castanet et al. 2000) : 0.4242 Observed compactness=0.77 Modeled compactness at the center=0.199 From 30 ° to 35 ° Modeled compactness=0.768 Modeled compactness at the periphery=0.994 Maximum likelihood:-Ln L=351.7913 Modeled compactness at the center=0.18 R/t (Currey & Alexander 1985) : 2.3684 AIC=711.5825 Modeled compactness at the periphery=0.955 r2=0.4515728 R/t (Currey & Alexander 1985) : 1.5612

206

CDI (Castanet et al. 2000) : 0.6405 Observed compactness=0.793 CDI (Castanet et al. 2000) : 0.4464 From 60 ° to 65 ° Modeled compactness=0.795 From 114 ° to 119 ° Maximum likelihood:-Ln L=309.9672 Modeled compactness at the center=0.376 Maximum likelihood:-Ln L=322.4112 AIC=627.9344 Modeled compactness at the periphery=0.983 AIC=652.8224 r2=0.5603214 R/t (Currey & Alexander 1985) : 2.0739 r2=0.4919693 Observed compactness=0.798 CDI (Castanet et al. 2000) : 0.4822 Observed compactness=0.764 Modeled compactness=0.801 From 90 ° to 95 ° Modeled compactness=0.764 Modeled compactness at the center=0.174 Maximum likelihood:-Ln L=286.0774 Modeled compactness at the center=0.196 Modeled compactness at the periphery=0.92 AIC=580.1547 Modeled compactness at the periphery=0.933 R/t (Currey & Alexander 1985) : 1.4558 r2=0.6293671 R/t (Currey & Alexander 1985) : 1.5138 CDI (Castanet et al. 2000) : 0.6869 Observed compactness=0.665 CDI (Castanet et al. 2000) : 0.6606 From 66 ° to 71 ° Modeled compactness=0.683 From 120 ° to 125 ° Maximum likelihood:-Ln L=297.0038 Modeled compactness at the center=0.211 Maximum likelihood:-Ln L=339.7931 AIC=602.0076 Modeled compactness at the periphery=0.978 AIC=687.5862 r2=0.5289236 R/t (Currey & Alexander 1985) : 2.5074 r2=0.4263514 Observed compactness=0.779 CDI (Castanet et al. 2000) : 0.3988 Observed compactness=0.783 Modeled compactness=0.783 From 96 ° to 101 ° Modeled compactness=0.786 Modeled compactness at the center=0.228 Maximum likelihood:-Ln L=256.6036 Modeled compactness at the center=0.45 Modeled compactness at the periphery=0.951 AIC=521.2072 Modeled compactness at the periphery=0.987 R/t (Currey & Alexander 1985) : 1.4574 r2=0.6272731 R/t (Currey & Alexander 1985) : 2.4497 CDI (Castanet et al. 2000) : 0.6862 Observed compactness=0.68 CDI (Castanet et al. 2000) : 0.4082 From 72 ° to 77 ° Modeled compactness=0.685 From 126 ° to 131 ° Maximum likelihood:-Ln L=370.478 Modeled compactness at the center=0.245 Maximum likelihood:-Ln L=243.043 AIC=748.9561 Modeled compactness at the periphery=0.981 AIC=494.0861 r2=0.4874541 R/t (Currey & Alexander 1985) : 2.609 r2=0.3625667 Observed compactness=0.737 CDI (Castanet et al. 2000) : 0.3833 Observed compactness=0.707 Modeled compactness=0.737 From 102 ° to 107 ° Modeled compactness=0.705 Modeled compactness at the center=0.328 Maximum likelihood:-Ln L=311.9335 Modeled compactness at the center=0.587 Modeled compactness at the periphery=0.821 AIC=631.8671 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 1.6857 r2=0.6596577 R/t (Currey & Alexander 1985) : 6.2966 CDI (Castanet et al. 2000) : 0.5932 Observed compactness=0.708 CDI (Castanet et al. 2000) : 0.1588 From 78 ° to 83 ° Modeled compactness=0.722 From 132 ° to 137 ° Maximum likelihood:-Ln L=352.2104 Modeled compactness at the center=0.168 Maximum likelihood:-Ln L=359.624 AIC=712.4209 Modeled compactness at the periphery=0.937 AIC=727.248 r2=0.391438 R/t (Currey & Alexander 1985) : 2.0044 r2=0.5504249 Observed compactness=0.818 CDI (Castanet et al. 2000) : 0.4989 Observed compactness=0.66 Modeled compactness=0.817 From 108 ° to 113 ° Modeled compactness=0.665 Modeled compactness at the center=0.353 Maximum likelihood:-Ln L=332.1686 Modeled compactness at the center=0.215 Modeled compactness at the periphery=0.928 AIC=672.3372 Modeled compactness at the periphery=0.859 R/t (Currey & Alexander 1985) : 1.1747 r2=0.4659083 R/t (Currey & Alexander 1985) : 1.7288 CDI (Castanet et al. 2000) : 0.8513 Observed compactness=0.771 CDI (Castanet et al. 2000) : 0.5784 From 84 ° to 89 ° Modeled compactness=0.772 From 138 ° to 143 ° Maximum likelihood:-Ln L=278.8611 Modeled compactness at the center=0.35 Maximum likelihood:-Ln L=313.253 AIC=565.7221 Modeled compactness at the periphery=0.987 AIC=634.5059 r2=0.6388373 R/t (Currey & Alexander 1985) : 2.24 r2=0.5675631

207

Observed compactness=0.704 CDI (Castanet et al. 2000) : 0.6447 Observed compactness=0.786 Modeled compactness=0.701 From 168 ° to 173 ° Modeled compactness=0.789 Modeled compactness at the center=0.187 Maximum likelihood:-Ln L=164.2717 Modeled compactness at the center=0.692 Modeled compactness at the periphery=0.927 AIC=336.5434 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 1.9373 r2=0.617439 R/t (Currey & Alexander 1985) : 5.637 CDI (Castanet et al. 2000) : 0.5162 Observed compactness=0.618 CDI (Castanet et al. 2000) : 0.1774 From 144 ° to 149 ° Modeled compactness=0.614 From 198 ° to 203 ° Maximum likelihood:-Ln L=262.1996 Modeled compactness at the center=0.097 Maximum likelihood:-Ln L=374.1083 AIC=532.3991 Modeled compactness at the periphery=0.875 AIC=756.2167 r2=0.2497238 R/t (Currey & Alexander 1985) : 2.1572 r2=0.1890909 Observed compactness=0.701 CDI (Castanet et al. 2000) : 0.4636 Observed compactness=0.779 Modeled compactness=0.704 From 174 ° to 179 ° Modeled compactness=0.782 Modeled compactness at the center=0.602 Maximum likelihood:-Ln L=151.7753 Modeled compactness at the center=0.7 Modeled compactness at the periphery=1. AIC=311.5506 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 7.1169 r2=0.3507972 R/t (Currey & Alexander 1985) : 6.5733 CDI (Castanet et al. 2000) : 0.1405 Observed compactness=0.529 CDI (Castanet et al. 2000) : 0.1521 From 150 ° to 155 ° Modeled compactness=0.517 From 204 ° to 209 ° Maximum likelihood:-Ln L=250.1291 Modeled compactness at the center=0.033 Maximum likelihood:-Ln L=254.8957 AIC=508.2582 Modeled compactness at the periphery=0.635 AIC=517.7915 r2=0.4742818 R/t (Currey & Alexander 1985) : 1.599 r2=0.2102492 Observed compactness=0.607 CDI (Castanet et al. 2000) : 0.6254 Observed compactness=0.762 Modeled compactness=0.613 From 180 ° to 185 ° Modeled compactness=0.764 Modeled compactness at the center=0.307 Maximum likelihood:-Ln L=216.9165 Modeled compactness at the center=0.702 Modeled compactness at the periphery=0.86 AIC=441.833 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 3.5584 r2=0.3511049 R/t (Currey & Alexander 1985) : 9.0199 CDI (Castanet et al. 2000) : 0.281 Observed compactness=0.837 CDI (Castanet et al. 2000) : 0.1109 From 156 ° to 161 ° Modeled compactness=0.831 From 210 ° to 215 ° Maximum likelihood:-Ln L=321.6418 Modeled compactness at the center=0.595 Maximum likelihood:-Ln L=382.4985 AIC=651.2836 Modeled compactness at the periphery=0.985 AIC=772.9969 r2=0.4474721 R/t (Currey & Alexander 1985) : 2.578 r2=0.1716028 Observed compactness=0.693 CDI (Castanet et al. 2000) : 0.3879 Observed compactness=0.681 Modeled compactness=0.692 From 186 ° to 191 ° Modeled compactness=0.677 Modeled compactness at the center=0.186 Maximum likelihood:-Ln L=324.1383 Modeled compactness at the center=0.585 Modeled compactness at the periphery=0.911 AIC=656.2766 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 1.8739 r2=0.260347 R/t (Currey & Alexander 1985) : 8.2693 CDI (Castanet et al. 2000) : 0.5336 Observed compactness=0.81 CDI (Castanet et al. 2000) : 0.1209 From 162 ° to 167 ° Modeled compactness=0.808 From 216 ° to 221 ° Maximum likelihood:-Ln L=246.1299 Modeled compactness at the center=0.666 Maximum likelihood:-Ln L=244.7166 AIC=500.2598 Modeled compactness at the periphery=0.991 AIC=497.4332 r2=0.415536 R/t (Currey & Alexander 1985) : 3.9294 r2=0.1344952 Observed compactness=0.645 CDI (Castanet et al. 2000) : 0.2545 Observed compactness=0.543 Modeled compactness=0.644 From 192 ° to 197 ° Modeled compactness=0.54 Modeled compactness at the center=0.273 Maximum likelihood:-Ln L=351.0436 Modeled compactness at the center=0.505 Modeled compactness at the periphery=0.8 AIC=710.0872 Modeled compactness at the periphery=0.996 R/t (Currey & Alexander 1985) : 1.5512 r2=0.2519763 R/t (Currey & Alexander 1985) : 27.7133

208

CDI (Castanet et al. 2000) : 0.0361 Observed compactness=0.702 CDI (Castanet et al. 2000) : 0.402 From 222 ° to 227 ° Modeled compactness=0.698 From 276 ° to 281 ° Maximum likelihood:-Ln L=365.9369 Modeled compactness at the center=0.528 Maximum likelihood:-Ln L=263.532 AIC=739.8737 Modeled compactness at the periphery=0.952 AIC=535.064 r2=0.1635616 R/t (Currey & Alexander 1985) : 4.7932 r2=0.6983789 Observed compactness=0.672 CDI (Castanet et al. 2000) : 0.2086 Observed compactness=0.807 Modeled compactness=0.67 From 252 ° to 257 ° Modeled compactness=0.805 Modeled compactness at the center=0.444 Maximum likelihood:-Ln L=128.8759 Modeled compactness at the center=0.004 Modeled compactness at the periphery=0.715 AIC=265.7518 Modeled compactness at the periphery=0.84 R/t (Currey & Alexander 1985) : 1.6809 r2=0.0391651 R/t (Currey & Alexander 1985) : 1.2372 CDI (Castanet et al. 2000) : 0.5949 Observed compactness=0.611 CDI (Castanet et al. 2000) : 0.8083 From 228 ° to 233 ° Modeled compactness=0.611 From 282 ° to 287 ° Maximum likelihood:-Ln L=136.9606 Modeled compactness at the center=0.602 Maximum likelihood:-Ln L=296.9389 AIC=281.9211 Modeled compactness at the periphery=0.615 AIC=601.8777 r2=0.4980223 R/t (Currey & Alexander 1985) : 0.8848 r2=0.17427 Observed compactness=0.41 CDI (Castanet et al. 2000) : 1.1301 Observed compactness=0.766 Modeled compactness=0.402 From 258 ° to 263 ° Modeled compactness=0.768 Modeled compactness at the center=0.26 Maximum likelihood:-Ln L=105.4834 Modeled compactness at the center=0.683 Modeled compactness at the periphery=1. AIC=218.9669 Modeled compactness at the periphery=0.959 R/t (Currey & Alexander 1985) : 9.7369 r2=0.2264045 R/t (Currey & Alexander 1985) : 5.8679 CDI (Castanet et al. 2000) : 0.1027 Observed compactness=0.514 CDI (Castanet et al. 2000) : 0.1704 From 234 ° to 239 ° Modeled compactness=0.508 From 288 ° to 293 ° Maximum likelihood:-Ln L=204.5007 Modeled compactness at the center=0.393 Maximum likelihood:-Ln L=265.7792 AIC=417.0015 Modeled compactness at the periphery=0.901 AIC=539.5584 r2=0.1125672 R/t (Currey & Alexander 1985) : 10.3595 r2=0.7008185 Observed compactness=0.62 CDI (Castanet et al. 2000) : 0.0965 Observed compactness=0.788 Modeled compactness=0.616 From 264 ° to 269 ° Modeled compactness=0.787 Modeled compactness at the center=0.566 Maximum likelihood:-Ln L=119.5569 Modeled compactness at the center=0. Modeled compactness at the periphery=1. AIC=247.1137 Modeled compactness at the periphery=0.811 R/t (Currey & Alexander 1985) : 15.884 r2=0.5033083 R/t (Currey & Alexander 1985) : 1.2035 CDI (Castanet et al. 2000) : 0.063 Observed compactness=0.507 CDI (Castanet et al. 2000) : 0.8309 From 240 ° to 245 ° Modeled compactness=0.517 From 294 ° to 299 ° Maximum likelihood:-Ln L=153.5051 Modeled compactness at the center=0. Maximum likelihood:-Ln L=363.5078 AIC=315.0101 Modeled compactness at the periphery=0.56 AIC=735.0157 r2=0.0737018 R/t (Currey & Alexander 1985) : 1.3731 r2=0.4113331 Observed compactness=0.698 CDI (Castanet et al. 2000) : 0.7283 Observed compactness=0.715 Modeled compactness=0.693 From 270 ° to 275 ° Modeled compactness=0.712 Modeled compactness at the center=0.667 Maximum likelihood:-Ln L=421.931 Modeled compactness at the center=0.245 Modeled compactness at the periphery=1. AIC=851.862 Modeled compactness at the periphery=0.893 R/t (Currey & Alexander 1985) : 25.3856 r2=0.3665841 R/t (Currey & Alexander 1985) : 1.5749 CDI (Castanet et al. 2000) : 0.0394 Observed compactness=0.748 CDI (Castanet et al. 2000) : 0.635 From 246 ° to 251 ° Modeled compactness=0.749 From 300 ° to 305 ° Maximum likelihood:-Ln L=128.2353 Modeled compactness at the center=0.409 Maximum likelihood:-Ln L=461.0331 AIC=264.4706 Modeled compactness at the periphery=0.956 AIC=930.0662 r2=0.2319461 R/t (Currey & Alexander 1985) : 2.4874 r2=0.4697965

209

Observed compactness=0.656 CDI (Castanet et al. 2000) : 0.8925 Observed compactness=0.778 Modeled compactness=0.651 From 330 ° to 335 ° Modeled compactness=0.77 Modeled compactness at the center=0.18 Maximum likelihood:-Ln L=405.725 Modeled compactness at the center=0.266 Modeled compactness at the periphery=0.933 AIC=819.4501 Modeled compactness at the periphery=0.953 R/t (Currey & Alexander 1985) : 2.4722 r2=0.2260041 R/t (Currey & Alexander 1985) : 1.7069 CDI (Castanet et al. 2000) : 0.4045 Observed compactness=0.828 CDI (Castanet et al. 2000) : 0.5859 From 306 ° to 311 ° Modeled compactness=0.829 Maximum likelihood:-Ln L=350.6226 Modeled compactness at the center=0.655 AIC=709.2452 Modeled compactness at the periphery=0.907 r2=0.253262 R/t (Currey & Alexander 1985) : 1.6654 NSM.005.GF.045.048-A Observed compactness=0.74 CDI (Castanet et al. 2000) : 0.6005 Global analysis Modeled compactness=0.735 From 336 ° to 341 ° Maximum likelihood:-Ln L=1401.263 Modeled compactness at the center=0.605 Maximum likelihood:-Ln L=388.2525 AIC=2810.527 Modeled compactness at the periphery=0.995 AIC=784.5049 r2=0.9185663 R/t (Currey & Alexander 1985) : 5.3187 r2=0.2961356 Observed compactness=0.62 CDI (Castanet et al. 2000) : 0.188 Observed compactness=0.796 Modeled compactness=0.62 From 312 ° to 317 ° Modeled compactness=0.797 Modeled compactness at the center=0.29 Maximum likelihood:-Ln L=342.7057 Modeled compactness at the center=0.562 Modeled compactness at the periphery=0.972 AIC=693.4114 Modeled compactness at the periphery=0.903 R/t (Currey & Alexander 1985) : 3.523 r2=0.5129865 R/t (Currey & Alexander 1985) : 1.7849 CDI (Castanet et al. 2000) : 0.2839 Observed compactness=0.729 CDI (Castanet et al. 2000) : 0.5603

Modeled compactness=0.719 From 342 ° to 347 ° Angular analysis Modeled compactness at the center=0.358 Maximum likelihood:-Ln L=352.7352 From 0 ° to 5 ° Modeled compactness at the periphery=0.937 AIC=713.4703 Maximum likelihood:-Ln L=552.4584 R/t (Currey & Alexander 1985) : 2.4996 r2=0.5632666 AIC=1112.917 CDI (Castanet et al. 2000) : 0.4001 Observed compactness=0.842 r2=0.2712496 From 318 ° to 323 ° Modeled compactness=0.84 Observed compactness=0.619 Maximum likelihood:-Ln L=398.483 Modeled compactness at the center=0.41 Modeled compactness=0.613 AIC=804.966 Modeled compactness at the periphery=0.954 Modeled compactness at the center=0.271 r2=0.4772329 R/t (Currey & Alexander 1985) : 1.1309 Modeled compactness at the periphery=0.775 Observed compactness=0.776 CDI (Castanet et al. 2000) : 0.8842 R/t (Currey & Alexander 1985) : 1.8119 Modeled compactness=0.773 From 348 ° to 353 ° CDI (Castanet et al. 2000) : 0.5519 Modeled compactness at the center=0.31 Maximum likelihood:-Ln L=296.8634 From 6 ° to 11 ° Modeled compactness at the periphery=0.923 AIC=601.7268 Maximum likelihood:-Ln L=545.0154 R/t (Currey & Alexander 1985) : 1.3335 r2=0.5342621 AIC=1098.031 CDI (Castanet et al. 2000) : 0.7499 Observed compactness=0.811 r2=0.1724973 From 324 ° to 329 ° Modeled compactness=0.814 Observed compactness=0.483 Maximum likelihood:-Ln L=340.243 Modeled compactness at the center=0.463 Modeled compactness=0.482 AIC=688.4861 Modeled compactness at the periphery=0.976 Modeled compactness at the center=0.287 r2=0.4285414 R/t (Currey & Alexander 1985) : 2.0748 Modeled compactness at the periphery=0.584 Observed compactness=0.814 CDI (Castanet et al. 2000) : 0.482 R/t (Currey & Alexander 1985) : 2.6311 Modeled compactness=0.814 From 354 ° to 359 ° CDI (Castanet et al. 2000) : 0.3801 Modeled compactness at the center=0.429 Maximum likelihood:-Ln L=397.3978 From 12 ° to 17 ° Modeled compactness at the periphery=0.93 AIC=802.7957 Maximum likelihood:-Ln L=325.4704 R/t (Currey & Alexander 1985) : 1.1204 r2=0.5144731

210

AIC=658.9407 Modeled compactness at the periphery=1. AIC=507.0519 r2=0.4150805 R/t (Currey & Alexander 1985) : 5.956 r2=0.5470418 Observed compactness=0.575 CDI (Castanet et al. 2000) : 0.1679 Observed compactness=0.598 Modeled compactness=0.574 From 42 ° to 47 ° Modeled compactness=0.6 Modeled compactness at the center=0.061 Maximum likelihood:-Ln L=270.1844 Modeled compactness at the center=0.282 Modeled compactness at the periphery=0.641 AIC=548.3688 Modeled compactness at the periphery=0.99 R/t (Currey & Alexander 1985) : 1.4617 r2=0.4826965 R/t (Currey & Alexander 1985) : 3.7761 CDI (Castanet et al. 2000) : 0.6841 Observed compactness=0.636 CDI (Castanet et al. 2000) : 0.2648 From 18 ° to 23 ° Modeled compactness=0.631 From 72 ° to 77 ° Maximum likelihood:-Ln L=510.9523 Modeled compactness at the center=0.42 Maximum likelihood:-Ln L=262.2918 AIC=1029.905 Modeled compactness at the periphery=1. AIC=532.5836 r2=0.326615 R/t (Currey & Alexander 1985) : 4.8315 r2=0.512732 Observed compactness=0.62 CDI (Castanet et al. 2000) : 0.207 Observed compactness=0.582 Modeled compactness=0.616 From 48 ° to 53 ° Modeled compactness=0.581 Modeled compactness at the center=0.398 Maximum likelihood:-Ln L=257.697 Modeled compactness at the center=0.13 Modeled compactness at the periphery=0.952 AIC=523.394 Modeled compactness at the periphery=0.921 R/t (Currey & Alexander 1985) : 4.7644 r2=0.4718529 R/t (Currey & Alexander 1985) : 2.928 CDI (Castanet et al. 2000) : 0.2099 Observed compactness=0.558 CDI (Castanet et al. 2000) : 0.3415 From 24 ° to 29 ° Modeled compactness=0.556 From 78 ° to 83 ° Maximum likelihood:-Ln L=454.863 Modeled compactness at the center=0.369 Maximum likelihood:-Ln L=178.3605 AIC=917.726 Modeled compactness at the periphery=0.999 AIC=364.7209 r2=0.3605474 R/t (Currey & Alexander 1985) : 6.0599 r2=0.4556026 Observed compactness=0.517 CDI (Castanet et al. 2000) : 0.165 Observed compactness=0.538 Modeled compactness=0.515 From 54 ° to 59 ° Modeled compactness=0.534 Modeled compactness at the center=0.371 Maximum likelihood:-Ln L=255.8048 Modeled compactness at the center=0.338 Modeled compactness at the periphery=1. AIC=519.6095 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 8.0105 r2=0.5227179 R/t (Currey & Alexander 1985) : 6.0748 CDI (Castanet et al. 2000) : 0.1248 Observed compactness=0.598 CDI (Castanet et al. 2000) : 0.1646 From 30 ° to 35 ° Modeled compactness=0.597 From 84 ° to 89 ° Maximum likelihood:-Ln L=341.7783 Modeled compactness at the center=0.347 Maximum likelihood:-Ln L=325.1038 AIC=691.5566 Modeled compactness at the periphery=1. AIC=658.2075 r2=0.3625239 R/t (Currey & Alexander 1985) : 4.5654 r2=0.3312057 Observed compactness=0.645 CDI (Castanet et al. 2000) : 0.219 Observed compactness=0.656 Modeled compactness=0.64 From 60 ° to 65 ° Modeled compactness=0.657 Modeled compactness at the center=0.493 Maximum likelihood:-Ln L=328.789 Modeled compactness at the center=0.461 Modeled compactness at the periphery=1. AIC=665.578 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 6.087 r2=0.4065522 R/t (Currey & Alexander 1985) : 4.8352 CDI (Castanet et al. 2000) : 0.1643 Observed compactness=0.554 CDI (Castanet et al. 2000) : 0.2068 From 36 ° to 41 ° Modeled compactness=0.565 From 90 ° to 95 ° Maximum likelihood:-Ln L=304.2775 Modeled compactness at the center=0.356 Maximum likelihood:-Ln L=471.1685 AIC=616.5549 Modeled compactness at the periphery=1. AIC=950.3371 r2=0.3227752 R/t (Currey & Alexander 1985) : 5.4966 r2=0.259166 Observed compactness=0.654 CDI (Castanet et al. 2000) : 0.1819 Observed compactness=0.696 Modeled compactness=0.649 From 66 ° to 71 ° Modeled compactness=0.696 Modeled compactness at the center=0.5 Maximum likelihood:-Ln L=249.5259 Modeled compactness at the center=0.588

211

Modeled compactness at the periphery=0.999 AIC=1441.288 Modeled compactness at the periphery=0.992 R/t (Currey & Alexander 1985) : 6.9213 r2=0.3510516 R/t (Currey & Alexander 1985) : 4.3109 CDI (Castanet et al. 2000) : 0.1445 Observed compactness=0.677 CDI (Castanet et al. 2000) : 0.232 From 96 ° to 101 ° Modeled compactness=0.682 From 150 ° to 155 ° Maximum likelihood:-Ln L=673.1083 Modeled compactness at the center=0.416 Maximum likelihood:-Ln L=521.2258 AIC=1354.217 Modeled compactness at the periphery=0.987 AIC=1050.452 r2=0.4034048 R/t (Currey & Alexander 1985) : 3.6325 r2=0.6076772 Observed compactness=0.664 CDI (Castanet et al. 2000) : 0.2753 Observed compactness=0.682 Modeled compactness=0.66 From 126 ° to 131 ° Modeled compactness=0.683 Modeled compactness at the center=0.41 Maximum likelihood:-Ln L=543.1857 Modeled compactness at the center=0.089 Modeled compactness at the periphery=1. AIC=1094.371 Modeled compactness at the periphery=0.943 R/t (Currey & Alexander 1985) : 4.0573 r2=0.313007 R/t (Currey & Alexander 1985) : 1.9839 CDI (Castanet et al. 2000) : 0.2465 Observed compactness=0.71 CDI (Castanet et al. 2000) : 0.5041 From 102 ° to 107 ° Modeled compactness=0.708 From 156 ° to 161 ° Maximum likelihood:-Ln L=494.5631 Modeled compactness at the center=0.531 Maximum likelihood:-Ln L=568.7042 AIC=997.1262 Modeled compactness at the periphery=0.996 AIC=1145.408 r2=0.4738354 R/t (Currey & Alexander 1985) : 4.5854 r2=0.6691256 Observed compactness=0.741 CDI (Castanet et al. 2000) : 0.2181 Observed compactness=0.687 Modeled compactness=0.752 From 132 ° to 137 ° Modeled compactness=0.687 Modeled compactness at the center=0.452 Maximum likelihood:-Ln L=609.4149 Modeled compactness at the center=0.142 Modeled compactness at the periphery=0.998 AIC=1226.83 Modeled compactness at the periphery=0.808 R/t (Currey & Alexander 1985) : 2.942 r2=0.4054662 R/t (Currey & Alexander 1985) : 1.7242 CDI (Castanet et al. 2000) : 0.3399 Observed compactness=0.695 CDI (Castanet et al. 2000) : 0.58 From 108 ° to 113 ° Modeled compactness=0.696 From 162 ° to 167 ° Maximum likelihood:-Ln L=414.5304 Modeled compactness at the center=0.508 Maximum likelihood:-Ln L=411.7925 AIC=837.0607 Modeled compactness at the periphery=0.939 AIC=831.5851 r2=0.5302641 R/t (Currey & Alexander 1985) : 3.9685 r2=0.7576448 Observed compactness=0.692 CDI (Castanet et al. 2000) : 0.252 Observed compactness=0.721 Modeled compactness=0.695 From 138 ° to 143 ° Modeled compactness=0.723 Modeled compactness at the center=0.367 Maximum likelihood:-Ln L=627.9963 Modeled compactness at the center=0.088 Modeled compactness at the periphery=0.973 AIC=1263.993 Modeled compactness at the periphery=0.979 R/t (Currey & Alexander 1985) : 3.0549 r2=0.3397857 R/t (Currey & Alexander 1985) : 1.9667 CDI (Castanet et al. 2000) : 0.3273 Observed compactness=0.7 CDI (Castanet et al. 2000) : 0.5085 From 114 ° to 119 ° Modeled compactness=0.7 From 168 ° to 173 ° Maximum likelihood:-Ln L=577.7534 Modeled compactness at the center=0.537 Maximum likelihood:-Ln L=479.8295 AIC=1163.507 Modeled compactness at the periphery=0.934 AIC=967.659 r2=0.4903012 R/t (Currey & Alexander 1985) : 4.2717 r2=0.7435622 Observed compactness=0.696 CDI (Castanet et al. 2000) : 0.2341 Observed compactness=0.634 Modeled compactness=0.7 From 144 ° to 149 ° Modeled compactness=0.641 Modeled compactness at the center=0.334 Maximum likelihood:-Ln L=418.073 Modeled compactness at the center=0.097 Modeled compactness at the periphery=0.996 AIC=844.146 Modeled compactness at the periphery=0.944 R/t (Currey & Alexander 1985) : 2.9141 r2=0.4534768 R/t (Currey & Alexander 1985) : 2.3626 CDI (Castanet et al. 2000) : 0.3432 Observed compactness=0.693 CDI (Castanet et al. 2000) : 0.4233 From 120 ° to 125 ° Modeled compactness=0.7 From 174 ° to 179 ° Maximum likelihood:-Ln L=716.644 Modeled compactness at the center=0.499 Maximum likelihood:-Ln L=416.4265

212

AIC=840.8529 Modeled compactness at the periphery=0.959 AIC=765.1521 r2=0.5360303 R/t (Currey & Alexander 1985) : 3.0611 r2=0.5707255 Observed compactness=0.692 CDI (Castanet et al. 2000) : 0.3267 Observed compactness=0.593 Modeled compactness=0.682 From 204 ° to 209 ° Modeled compactness=0.584 Modeled compactness at the center=0.206 Maximum likelihood:-Ln L=394.5204 Modeled compactness at the center=0.04 Modeled compactness at the periphery=0.984 AIC=797.0409 Modeled compactness at the periphery=0.961 R/t (Currey & Alexander 1985) : 2.5266 r2=0.5779446 R/t (Currey & Alexander 1985) : 2.6779 CDI (Castanet et al. 2000) : 0.3958 Observed compactness=0.511 CDI (Castanet et al. 2000) : 0.3734 From 180 ° to 185 ° Modeled compactness=0.52 From 234 ° to 239 ° Maximum likelihood:-Ln L=338.8332 Modeled compactness at the center=0.073 Maximum likelihood:-Ln L=324.4619 AIC=685.6663 Modeled compactness at the periphery=0.886 AIC=656.9237 r2=0.7751175 R/t (Currey & Alexander 1985) : 2.946 r2=0.6089938 Observed compactness=0.545 CDI (Castanet et al. 2000) : 0.3394 Observed compactness=0.649 Modeled compactness=0.531 From 210 ° to 215 ° Modeled compactness=0.64 Modeled compactness at the center=0.067 Maximum likelihood:-Ln L=356.3388 Modeled compactness at the center=0.065 Modeled compactness at the periphery=0.977 AIC=720.6776 Modeled compactness at the periphery=0.987 R/t (Currey & Alexander 1985) : 3.2464 r2=0.5730503 R/t (Currey & Alexander 1985) : 2.466 CDI (Castanet et al. 2000) : 0.308 Observed compactness=0.607 CDI (Castanet et al. 2000) : 0.4055 From 186 ° to 191 ° Modeled compactness=0.604 From 240 ° to 245 ° Maximum likelihood:-Ln L=261.4638 Modeled compactness at the center=0.079 Maximum likelihood:-Ln L=272.3009 AIC=530.9276 Modeled compactness at the periphery=0.971 AIC=552.6017 r2=0.681944 R/t (Currey & Alexander 1985) : 2.6802 r2=0.61383 Observed compactness=0.535 CDI (Castanet et al. 2000) : 0.3731 Observed compactness=0.549 Modeled compactness=0.532 From 216 ° to 221 ° Modeled compactness=0.558 Modeled compactness at the center=0.152 Maximum likelihood:-Ln L=334.8067 Modeled compactness at the center=0.253 Modeled compactness at the periphery=0.994 AIC=677.6135 Modeled compactness at the periphery=0.945 R/t (Currey & Alexander 1985) : 3.753 r2=0.6354887 R/t (Currey & Alexander 1985) : 3.8849 CDI (Castanet et al. 2000) : 0.2665 Observed compactness=0.641 CDI (Castanet et al. 2000) : 0.2574 From 192 ° to 197 ° Modeled compactness=0.638 From 246 ° to 251 ° Maximum likelihood:-Ln L=309.1357 Modeled compactness at the center=0.129 Maximum likelihood:-Ln L=419.4615 AIC=626.2714 Modeled compactness at the periphery=0.922 AIC=846.923 r2=0.5173469 R/t (Currey & Alexander 1985) : 2.375 r2=0.63219 Observed compactness=0.54 CDI (Castanet et al. 2000) : 0.4211 Observed compactness=0.638 Modeled compactness=0.541 From 222 ° to 227 ° Modeled compactness=0.642 Modeled compactness at the center=0.08 Maximum likelihood:-Ln L=296.1797 Modeled compactness at the center=0.159 Modeled compactness at the periphery=0.969 AIC=600.3594 Modeled compactness at the periphery=0.963 R/t (Currey & Alexander 1985) : 3.1983 r2=0.6983596 R/t (Currey & Alexander 1985) : 2.6097 CDI (Castanet et al. 2000) : 0.3127 Observed compactness=0.667 CDI (Castanet et al. 2000) : 0.3832 From 198 ° to 203 ° Modeled compactness=0.663 From 252 ° to 257 ° Maximum likelihood:-Ln L=302.5576 Modeled compactness at the center=0.074 Maximum likelihood:-Ln L=328.4048 AIC=613.1152 Modeled compactness at the periphery=0.926 AIC=664.8095 r2=0.4023818 R/t (Currey & Alexander 1985) : 2.1524 r2=0.7596316 Observed compactness=0.589 CDI (Castanet et al. 2000) : 0.4646 Observed compactness=0.613 Modeled compactness=0.592 From 228 ° to 233 ° Modeled compactness=0.613 Modeled compactness at the center=0.159 Maximum likelihood:-Ln L=378.576 Modeled compactness at the center=0.242

213

Modeled compactness at the periphery=0.998 AIC=642.6483 Modeled compactness at the periphery=0.996 R/t (Currey & Alexander 1985) : 3.3966 r2=0.7763554 R/t (Currey & Alexander 1985) : 4.4042 CDI (Castanet et al. 2000) : 0.2944 Observed compactness=0.562 CDI (Castanet et al. 2000) : 0.2271 From 258 ° to 263 ° Modeled compactness=0.556 From 312 ° to 317 ° Maximum likelihood:-Ln L=426.0998 Modeled compactness at the center=0.093 Maximum likelihood:-Ln L=373.4923 AIC=860.1995 Modeled compactness at the periphery=0.974 AIC=754.9846 r2=0.5925592 R/t (Currey & Alexander 1985) : 3.1089 r2=0.6731681 Observed compactness=0.601 CDI (Castanet et al. 2000) : 0.3217 Observed compactness=0.571 Modeled compactness=0.598 From 288 ° to 293 ° Modeled compactness=0.554 Modeled compactness at the center=0.37 Maximum likelihood:-Ln L=257.0042 Modeled compactness at the center=0.153 Modeled compactness at the periphery=1. AIC=522.0083 Modeled compactness at the periphery=0.961 R/t (Currey & Alexander 1985) : 4.8822 r2=0.7349084 R/t (Currey & Alexander 1985) : 3.4219 CDI (Castanet et al. 2000) : 0.2048 Observed compactness=0.511 CDI (Castanet et al. 2000) : 0.2922 From 264 ° to 269 ° Modeled compactness=0.516 From 318 ° to 323 ° Maximum likelihood:-Ln L=269.6524 Modeled compactness at the center=0.186 Maximum likelihood:-Ln L=368.0974 AIC=547.3049 Modeled compactness at the periphery=1. AIC=744.1949 r2=0.8509964 R/t (Currey & Alexander 1985) : 4.2846 r2=0.6015299 Observed compactness=0.651 CDI (Castanet et al. 2000) : 0.2334 Observed compactness=0.564 Modeled compactness=0.654 From 294 ° to 299 ° Modeled compactness=0.557 Modeled compactness at the center=0.088 Maximum likelihood:-Ln L=320.3133 Modeled compactness at the center=0.256 Modeled compactness at the periphery=0.99 AIC=648.6266 Modeled compactness at the periphery=0.998 R/t (Currey & Alexander 1985) : 2.4464 r2=0.7687749 R/t (Currey & Alexander 1985) : 4.2392 CDI (Castanet et al. 2000) : 0.4088 Observed compactness=0.528 CDI (Castanet et al. 2000) : 0.2359 From 270 ° to 275 ° Modeled compactness=0.528 From 324 ° to 329 ° Maximum likelihood:-Ln L=291.5433 Modeled compactness at the center=0.106 Maximum likelihood:-Ln L=302.4968 AIC=591.0865 Modeled compactness at the periphery=0.991 AIC=612.9935 r2=0.6692699 R/t (Currey & Alexander 1985) : 3.5045 r2=0.7172874 Observed compactness=0.542 CDI (Castanet et al. 2000) : 0.2853 Observed compactness=0.539 Modeled compactness=0.548 From 300 ° to 305 ° Modeled compactness=0.542 Modeled compactness at the center=0.184 Maximum likelihood:-Ln L=258.2364 Modeled compactness at the center=0.19 Modeled compactness at the periphery=0.996 AIC=524.4727 Modeled compactness at the periphery=0.998 R/t (Currey & Alexander 1985) : 3.7722 r2=0.8119926 R/t (Currey & Alexander 1985) : 3.9086 CDI (Castanet et al. 2000) : 0.2651 Observed compactness=0.579 CDI (Castanet et al. 2000) : 0.2558 From 276 ° to 281 ° Modeled compactness=0.578 From 330 ° to 335 ° Maximum likelihood:-Ln L=314.4559 Modeled compactness at the center=0.116 Maximum likelihood:-Ln L=303.7583 AIC=636.9117 Modeled compactness at the periphery=0.997 AIC=615.5165 r2=0.754078 R/t (Currey & Alexander 1985) : 3.122 r2=0.7019112 Observed compactness=0.5 CDI (Castanet et al. 2000) : 0.3203 Observed compactness=0.569 Modeled compactness=0.519 From 306 ° to 311 ° Modeled compactness=0.562 Modeled compactness at the center=0.053 Maximum likelihood:-Ln L=352.7806 Modeled compactness at the center=0.205 Modeled compactness at the periphery=0.987 AIC=713.5612 Modeled compactness at the periphery=0.998 R/t (Currey & Alexander 1985) : 3.3264 r2=0.5442367 R/t (Currey & Alexander 1985) : 3.7609 CDI (Castanet et al. 2000) : 0.3006 Observed compactness=0.536 CDI (Castanet et al. 2000) : 0.2659 From 282 ° to 287 ° Modeled compactness=0.532 From 336 ° to 341 ° Maximum likelihood:-Ln L=317.3241 Modeled compactness at the center=0.23 Maximum likelihood:-Ln L=311.9655

214

AIC=631.9309 AIC=2571.186 Modeled compactness=0.677 r2=0.4550848 r2=0.8892297 Modeled compactness at the center=0.298 Observed compactness=0.539 Observed compactness=0.647 Modeled compactness at the periphery=0.793 Modeled compactness=0.563 Modeled compactness=0.647 R/t (Currey & Alexander 1985) : 1.9255 Modeled compactness at the center=0.358 Modeled compactness at the center=0.455 CDI (Castanet et al. 2000) : 0.5194 Modeled compactness at the periphery=1. Modeled compactness at the periphery=0.935 From 24 ° to 29 ° R/t (Currey & Alexander 1985) : 5.5941 R/t (Currey & Alexander 1985) : 4.9825 Maximum likelihood:-Ln L=654.9179 CDI (Castanet et al. 2000) : 0.1788 CDI (Castanet et al. 2000) : 0.2007 AIC=1317.836 From 342 ° to 347 ° r2=0.322003 Maximum likelihood:-Ln L=346.0305 Angular analysis Observed compactness=0.711 AIC=700.061 From 0 ° to 5 ° Modeled compactness=0.71 r2=0.3645998 Maximum likelihood:-Ln L=426.6209 Modeled compactness at the center=0.03 Observed compactness=0.609 AIC=861.2418 Modeled compactness at the periphery=0.762 Modeled compactness=0.611 r2=0.2518644 R/t (Currey & Alexander 1985) : 1.2971 Modeled compactness at the center=0.377 Observed compactness=0.675 CDI (Castanet et al. 2000) : 0.7709 Modeled compactness at the periphery=1. Modeled compactness=0.67 From 30 ° to 35 ° R/t (Currey & Alexander 1985) : 4.6727 Modeled compactness at the center=0.411 Maximum likelihood:-Ln L=360.2809 CDI (Castanet et al. 2000) : 0.214 Modeled compactness at the periphery=0.822 AIC=728.5618 From 348 ° to 353 ° R/t (Currey & Alexander 1985) : 2.5495 r2=0.5611838 Maximum likelihood:-Ln L=256.8848 CDI (Castanet et al. 2000) : 0.3922 Observed compactness=0.708 AIC=521.7697 From 6 ° to 11 ° Modeled compactness=0.707 r2=0.4182734 Maximum likelihood:-Ln L=839.7254 Modeled compactness at the center=0.241 Observed compactness=0.619 AIC=1687.451 Modeled compactness at the periphery=0.756 Modeled compactness=0.613 r2=0.0573384 R/t (Currey & Alexander 1985) : 1.4481 Modeled compactness at the center=0.357 Observed compactness=0.528 CDI (Castanet et al. 2000) : 0.6906 Modeled compactness at the periphery=0.991 Modeled compactness=0.527 From 36 ° to 41 ° R/t (Currey & Alexander 1985) : 4.3006 Modeled compactness at the center=0.495 Maximum likelihood:-Ln L=524.5679 CDI (Castanet et al. 2000) : 0.2325 Modeled compactness at the periphery=0.997 AIC=1057.136 From 354 ° to 359 ° R/t (Currey & Alexander 1985) : 32.0205 r2=0.1045603 Maximum likelihood:-Ln L=366.3684 CDI (Castanet et al. 2000) : 0.0312 Observed compactness=0.672 AIC=740.7368 From 12 ° to 17 ° Modeled compactness=0.671 r2=0.3808759 Maximum likelihood:-Ln L=454.1943 Modeled compactness at the center=0.626 Observed compactness=0.467 AIC=916.3885 Modeled compactness at the periphery=1. Modeled compactness=0.456 r2=0.0000177 R/t (Currey & Alexander 1985) : 16.073 Modeled compactness at the center=0.213 Observed compactness=0.672 CDI (Castanet et al. 2000) : 0.0622 Modeled compactness at the periphery=0.909 Modeled compactness=0.672 From 42 ° to 47 ° R/t (Currey & Alexander 1985) : 5.121 Modeled compactness at the center=0.693 Maximum likelihood:-Ln L=483.8105 CDI (Castanet et al. 2000) : 0.1953 Modeled compactness at the periphery=0.656 AIC=975.6211 R/t (Currey & Alexander 1985) : 0.8586 r2=0.373456 CDI (Castanet et al. 2000) : 1.1647 Observed compactness=0.559 From 18 ° to 23 ° Modeled compactness=0.554 RM.20.6711 Maximum likelihood:-Ln L=347.5655 Modeled compactness at the center=0.37 AIC=703.131 Modeled compactness at the periphery=1. Global analysis r2=0.2772282 R/t (Currey & Alexander 1985) : 6.1933 Maximum likelihood:-Ln L=1281.593 Observed compactness=0.68 CDI (Castanet et al. 2000) : 0.1615

215

From 48 ° to 53 ° Modeled compactness=0.668 From 102 ° to 107 ° Maximum likelihood:-Ln L=295.2823 Modeled compactness at the center=0.421 Maximum likelihood:-Ln L=455.4958 AIC=598.5646 Modeled compactness at the periphery=1. AIC=918.9917 r2=0.3788486 R/t (Currey & Alexander 1985) : 4.0622 r2=0.2150268 Observed compactness=0.627 CDI (Castanet et al. 2000) : 0.2462 Observed compactness=0.543 Modeled compactness=0.628 From 78 ° to 83 ° Modeled compactness=0.545 Modeled compactness at the center=0.497 Maximum likelihood:-Ln L=215.9067 Modeled compactness at the center=0.292 Modeled compactness at the periphery=0.997 AIC=439.8134 Modeled compactness at the periphery=0.881 R/t (Currey & Alexander 1985) : 6.9621 r2=0.1474316 R/t (Currey & Alexander 1985) : 4.9826 CDI (Castanet et al. 2000) : 0.1436 Observed compactness=0.62 CDI (Castanet et al. 2000) : 0.2007 From 54 ° to 59 ° Modeled compactness=0.617 From 108 ° to 113 ° Maximum likelihood:-Ln L=467.989 Modeled compactness at the center=0.522 Maximum likelihood:-Ln L=541.2701 AIC=943.978 Modeled compactness at the periphery=1. AIC=1090.54 r2=0.1370691 R/t (Currey & Alexander 1985) : 9.3105 r2=0.2121184 Observed compactness=0.611 CDI (Castanet et al. 2000) : 0.1074 Observed compactness=0.543 Modeled compactness=0.613 From 84 ° to 89 ° Modeled compactness=0.544 Modeled compactness at the center=0.544 Maximum likelihood:-Ln L=140.3553 Modeled compactness at the center=0.349 Modeled compactness at the periphery=1. AIC=288.7107 Modeled compactness at the periphery=0.824 R/t (Currey & Alexander 1985) : 12.5937 r2=0.2863215 R/t (Currey & Alexander 1985) : 4.2869 CDI (Castanet et al. 2000) : 0.0794 Observed compactness=0.601 CDI (Castanet et al. 2000) : 0.2333 From 60 ° to 65 ° Modeled compactness=0.603 From 114 ° to 119 ° Maximum likelihood:-Ln L=377.0262 Modeled compactness at the center=0.519 Maximum likelihood:-Ln L=564.9149 AIC=762.0525 Modeled compactness at the periphery=1. AIC=1137.83 r2=0.1132229 R/t (Currey & Alexander 1985) : 10.887 r2=0.0531343 Observed compactness=0.624 CDI (Castanet et al. 2000) : 0.0919 Observed compactness=0.642 Modeled compactness=0.626 From 90 ° to 95 ° Modeled compactness=0.642 Modeled compactness at the center=0.567 Maximum likelihood:-Ln L=284.2136 Modeled compactness at the center=0.388 Modeled compactness at the periphery=1. AIC=576.4272 Modeled compactness at the periphery=0.78 R/t (Currey & Alexander 1985) : 14.1639 r2=0.5197358 R/t (Currey & Alexander 1985) : 2.6549 CDI (Castanet et al. 2000) : 0.0706 Observed compactness=0.704 CDI (Castanet et al. 2000) : 0.3767 From 66 ° to 71 ° Modeled compactness=0.708 From 120 ° to 125 ° Maximum likelihood:-Ln L=287.6863 Modeled compactness at the center=0.202 Maximum likelihood:-Ln L=519.0966 AIC=583.3727 Modeled compactness at the periphery=1. AIC=1046.193 r2=0.2231627 R/t (Currey & Alexander 1985) : 2.4968 r2=0.1130899 Observed compactness=0.741 CDI (Castanet et al. 2000) : 0.4005 Observed compactness=0.635 Modeled compactness=0.738 From 96 ° to 101 ° Modeled compactness=0.632 Modeled compactness at the center=0.52 Maximum likelihood:-Ln L=677.7605 Modeled compactness at the center=0.513 Modeled compactness at the periphery=0.921 AIC=1363.521 Modeled compactness at the periphery=0.967 R/t (Currey & Alexander 1985) : 3.4862 r2=0.3229107 R/t (Currey & Alexander 1985) : 7.0636 CDI (Castanet et al. 2000) : 0.2868 Observed compactness=0.605 CDI (Castanet et al. 2000) : 0.1416 From 72 ° to 77 ° Modeled compactness=0.604 From 126 ° to 131 ° Maximum likelihood:-Ln L=174.533 Modeled compactness at the center=0.189 Maximum likelihood:-Ln L=323.6124 AIC=357.066 Modeled compactness at the periphery=0.771 AIC=655.2248 r2=0.3007966 R/t (Currey & Alexander 1985) : 2.1316 r2=0.29008 Observed compactness=0.67 CDI (Castanet et al. 2000) : 0.4691 Observed compactness=0.732

216

Modeled compactness=0.73 From 156 ° to 161 ° Modeled compactness=0.639 Modeled compactness at the center=0.52 Maximum likelihood:-Ln L=604.218 Modeled compactness at the center=0.353 Modeled compactness at the periphery=0.97 AIC=1216.436 Modeled compactness at the periphery=0.924 R/t (Currey & Alexander 1985) : 3.6752 r2=0.2735676 R/t (Currey & Alexander 1985) : 3.6892 CDI (Castanet et al. 2000) : 0.2721 Observed compactness=0.534 CDI (Castanet et al. 2000) : 0.2711 From 132 ° to 137 ° Modeled compactness=0.536 From 186 ° to 191 ° Maximum likelihood:-Ln L=362.9909 Modeled compactness at the center=0.426 Maximum likelihood:-Ln L=659.7802 AIC=733.9818 Modeled compactness at the periphery=1. AIC=1327.56 r2=0.4506924 R/t (Currey & Alexander 1985) : 9.3985 r2=0.4031647 Observed compactness=0.693 CDI (Castanet et al. 2000) : 0.1064 Observed compactness=0.607 Modeled compactness=0.696 From 162 ° to 167 ° Modeled compactness=0.606 Modeled compactness at the center=0.42 Maximum likelihood:-Ln L=788.5604 Modeled compactness at the center=0.372 Modeled compactness at the periphery=0.997 AIC=1585.121 Modeled compactness at the periphery=0.994 R/t (Currey & Alexander 1985) : 3.4935 r2=0.2162302 R/t (Currey & Alexander 1985) : 4.6443 CDI (Castanet et al. 2000) : 0.2862 Observed compactness=0.598 CDI (Castanet et al. 2000) : 0.2153 From 138 ° to 143 ° Modeled compactness=0.601 From 192 ° to 197 ° Maximum likelihood:-Ln L=362.958 Modeled compactness at the center=0.477 Maximum likelihood:-Ln L=493.6096 AIC=733.916 Modeled compactness at the periphery=1. AIC=995.2193 r2=0.3288536 R/t (Currey & Alexander 1985) : 7.741 r2=0.4931116 Observed compactness=0.794 CDI (Castanet et al. 2000) : 0.1292 Observed compactness=0.634 Modeled compactness=0.792 From 168 ° to 173 ° Modeled compactness=0.638 Modeled compactness at the center=0.615 Maximum likelihood:-Ln L=736.3983 Modeled compactness at the center=0.349 Modeled compactness at the periphery=1. AIC=1480.797 Modeled compactness at the periphery=0.992 R/t (Currey & Alexander 1985) : 3.7095 r2=0.1531007 R/t (Currey & Alexander 1985) : 3.7834 CDI (Castanet et al. 2000) : 0.2696 Observed compactness=0.63 CDI (Castanet et al. 2000) : 0.2643 From 144 ° to 149 ° Modeled compactness=0.63 From 198 ° to 203 ° Maximum likelihood:-Ln L=449.6747 Modeled compactness at the center=0.523 Maximum likelihood:-Ln L=360.7565 AIC=907.3494 Modeled compactness at the periphery=0.994 AIC=729.5131 r2=0.4018594 R/t (Currey & Alexander 1985) : 8.2084 r2=0.5065151 Observed compactness=0.757 CDI (Castanet et al. 2000) : 0.1218 Observed compactness=0.64 Modeled compactness=0.756 From 174 ° to 179 ° Modeled compactness=0.65 Modeled compactness at the center=0.37 Maximum likelihood:-Ln L=487.7671 Modeled compactness at the center=0.332 Modeled compactness at the periphery=0.999 AIC=983.5342 Modeled compactness at the periphery=0.994 R/t (Currey & Alexander 1985) : 2.5673 r2=0.2031365 R/t (Currey & Alexander 1985) : 3.4855 CDI (Castanet et al. 2000) : 0.3895 Observed compactness=0.608 CDI (Castanet et al. 2000) : 0.2869 From 150 ° to 155 ° Modeled compactness=0.613 From 204 ° to 209 ° Maximum likelihood:-Ln L=466.2702 Modeled compactness at the center=0.528 Maximum likelihood:-Ln L=408.7584 AIC=940.5403 Modeled compactness at the periphery=0.983 AIC=825.5167 r2=0.2771581 R/t (Currey & Alexander 1985) : 9.9865 r2=0.1897005 Observed compactness=0.722 CDI (Castanet et al. 2000) : 0.1001 Observed compactness=0.567 Modeled compactness=0.73 From 180 ° to 185 ° Modeled compactness=0.577 Modeled compactness at the center=0.536 Maximum likelihood:-Ln L=697.4952 Modeled compactness at the center=0.475 Modeled compactness at the periphery=0.998 AIC=1402.99 Modeled compactness at the periphery=0.966 R/t (Currey & Alexander 1985) : 4.0987 r2=0.3109651 R/t (Currey & Alexander 1985) : 9.651 CDI (Castanet et al. 2000) : 0.244 Observed compactness=0.645 CDI (Castanet et al. 2000) : 0.1036

217

From 210 ° to 215 ° Modeled compactness=0.718 From 264 ° to 269 ° Maximum likelihood:-Ln L=443.4794 Modeled compactness at the center=0.388 Maximum likelihood:-Ln L=643.2067 AIC=894.9589 Modeled compactness at the periphery=0.97 AIC=1294.413 r2=0.2707581 R/t (Currey & Alexander 1985) : 2.866 r2=0.2260059 Observed compactness=0.651 CDI (Castanet et al. 2000) : 0.3489 Observed compactness=0.548 Modeled compactness=0.648 From 240 ° to 245 ° Modeled compactness=0.553 Modeled compactness at the center=0.492 Maximum likelihood:-Ln L=811.7532 Modeled compactness at the center=0.446 Modeled compactness at the periphery=0.975 AIC=1631.506 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 5.791 r2=0.3466995 R/t (Currey & Alexander 1985) : 9.6755 CDI (Castanet et al. 2000) : 0.1727 Observed compactness=0.6 CDI (Castanet et al. 2000) : 0.1034 From 216 ° to 221 ° Modeled compactness=0.601 From 270 ° to 275 ° Maximum likelihood:-Ln L=413.58 Modeled compactness at the center=0.312 Maximum likelihood:-Ln L=731.0139 AIC=835.1599 Modeled compactness at the periphery=0.958 AIC=1470.028 r2=0.5397993 R/t (Currey & Alexander 1985) : 3.9895 r2=0.2801072 Observed compactness=0.587 CDI (Castanet et al. 2000) : 0.2507 Observed compactness=0.614 Modeled compactness=0.583 From 246 ° to 251 ° Modeled compactness=0.614 Modeled compactness at the center=0.234 Maximum likelihood:-Ln L=458.9071 Modeled compactness at the center=0.368 Modeled compactness at the periphery=0.941 AIC=925.8142 Modeled compactness at the periphery=0.917 R/t (Currey & Alexander 1985) : 3.4066 r2=0.1074869 R/t (Currey & Alexander 1985) : 4.391 CDI (Castanet et al. 2000) : 0.2935 Observed compactness=0.79 CDI (Castanet et al. 2000) : 0.2277 From 222 ° to 227 ° Modeled compactness=0.791 From 276 ° to 281 ° Maximum likelihood:-Ln L=571.8121 Modeled compactness at the center=0.713 Maximum likelihood:-Ln L=532.0542 AIC=1151.624 Modeled compactness at the periphery=0.995 AIC=1072.108 r2=0.3213812 R/t (Currey & Alexander 1985) : 6.6589 r2=0.4925592 Observed compactness=0.625 CDI (Castanet et al. 2000) : 0.1502 Observed compactness=0.662 Modeled compactness=0.622 From 252 ° to 257 ° Modeled compactness=0.66 Modeled compactness at the center=0.355 Maximum likelihood:-Ln L=704.0161 Modeled compactness at the center=0.355 Modeled compactness at the periphery=0.905 AIC=1416.032 Modeled compactness at the periphery=0.998 R/t (Currey & Alexander 1985) : 4.0476 r2=0.1117928 R/t (Currey & Alexander 1985) : 3.5446 CDI (Castanet et al. 2000) : 0.2471 Observed compactness=0.736 CDI (Castanet et al. 2000) : 0.2821 From 228 ° to 233 ° Modeled compactness=0.736 From 282 ° to 287 ° Maximum likelihood:-Ln L=576.9867 Modeled compactness at the center=0.666 Maximum likelihood:-Ln L=323.3945 AIC=1161.973 Modeled compactness at the periphery=1. AIC=654.7891 r2=0.1412045 R/t (Currey & Alexander 1985) : 9.0033 r2=0.5764403 Observed compactness=0.679 CDI (Castanet et al. 2000) : 0.1111 Observed compactness=0.697 Modeled compactness=0.676 From 258 ° to 263 ° Modeled compactness=0.705 Modeled compactness at the center=0.618 Maximum likelihood:-Ln L=656.0421 Modeled compactness at the center=0.409 Modeled compactness at the periphery=1. AIC=1320.084 Modeled compactness at the periphery=0.991 R/t (Currey & Alexander 1985) : 12.4015 r2=0.1798262 R/t (Currey & Alexander 1985) : 3.2789 CDI (Castanet et al. 2000) : 0.0806 Observed compactness=0.625 CDI (Castanet et al. 2000) : 0.305 From 234 ° to 239 ° Modeled compactness=0.623 From 288 ° to 293 ° Maximum likelihood:-Ln L=492.4121 Modeled compactness at the center=0.524 Maximum likelihood:-Ln L=525.6155 AIC=992.8242 Modeled compactness at the periphery=1. AIC=1059.231 r2=0.4699966 R/t (Currey & Alexander 1985) : 9.0302 r2=0.4660768 Observed compactness=0.713 CDI (Castanet et al. 2000) : 0.1107 Observed compactness=0.688

218

Modeled compactness=0.689 From 318 ° to 323 ° Modeled compactness=0.453 Modeled compactness at the center=0.402 Maximum likelihood:-Ln L=529.9703 Modeled compactness at the center=0.211 Modeled compactness at the periphery=0.991 AIC=1067.941 Modeled compactness at the periphery=0.969 R/t (Currey & Alexander 1985) : 3.4306 r2=0.1634617 R/t (Currey & Alexander 1985) : 5.7242 CDI (Castanet et al. 2000) : 0.2915 Observed compactness=0.602 CDI (Castanet et al. 2000) : 0.1747 From 294 ° to 299 ° Modeled compactness=0.603 From 348 ° to 353 ° Maximum likelihood:-Ln L=619.8835 Modeled compactness at the center=0.381 Maximum likelihood:-Ln L=217.7592 AIC=1247.767 Modeled compactness at the periphery=0.927 AIC=443.5183 r2=0.4226294 R/t (Currey & Alexander 1985) : 4.8721 r2=0.2602024 Observed compactness=0.658 CDI (Castanet et al. 2000) : 0.2052 Observed compactness=0.574 Modeled compactness=0.662 From 324 ° to 329 ° Modeled compactness=0.569 Modeled compactness at the center=0.443 Maximum likelihood:-Ln L=370.705 Modeled compactness at the center=0.368 Modeled compactness at the periphery=0.917 AIC=749.41 Modeled compactness at the periphery=0.934 R/t (Currey & Alexander 1985) : 3.6981 r2=0.1651559 R/t (Currey & Alexander 1985) : 5.608 CDI (Castanet et al. 2000) : 0.2704 Observed compactness=0.689 CDI (Castanet et al. 2000) : 0.1783 From 300 ° to 305 ° Modeled compactness=0.688 From 354 ° to 359 ° Maximum likelihood:-Ln L=531.7801 Modeled compactness at the center=0.43 Maximum likelihood:-Ln L=140.6803 AIC=1071.56 Modeled compactness at the periphery=0.859 AIC=289.3606 r2=0.1549967 R/t (Currey & Alexander 1985) : 3.1822 r2=0.3186498 Observed compactness=0.599 CDI (Castanet et al. 2000) : 0.3142 Observed compactness=0.629 Modeled compactness=0.596 From 330 ° to 335 ° Modeled compactness=0.62 Modeled compactness at the center=0.468 Maximum likelihood:-Ln L=176.3907 Modeled compactness at the center=0.447 Modeled compactness at the periphery=0.903 AIC=360.7815 Modeled compactness at the periphery=0.772 R/t (Currey & Alexander 1985) : 8.1421 r2=0.6092207 R/t (Currey & Alexander 1985) : 3.1177 CDI (Castanet et al. 2000) : 0.1228 Observed compactness=0.681 CDI (Castanet et al. 2000) : 0.3207 From 306 ° to 311 ° Modeled compactness=0.679 Maximum likelihood:-Ln L=560.941 Modeled compactness at the center=0.054 Platypus (MVZ M35370-70863) AIC=1129.882 Modeled compactness at the periphery=0.883 http://www.morphosource.org/Detail/MediaDetail/S r2=0.216242 R/t (Currey & Alexander 1985) : 1.7403 how/media_id/35370 Observed compactness=0.663 CDI (Castanet et al. 2000) : 0.5746 Modeled compactness=0.659 From 336 ° to 341 ° Global analysis Modeled compactness at the center=0.532 Maximum likelihood:-Ln L=200.8071 Maximum likelihood:-Ln L=1884.098 Modeled compactness at the periphery=0.984 AIC=409.6141 AIC=3776.195 R/t (Currey & Alexander 1985) : 6.5804 r2=0.6430156 r2=0.8521959 CDI (Castanet et al. 2000) : 0.152 Observed compactness=0.51 Observed compactness=0.736 From 312 ° to 317 ° Modeled compactness=0.507 Modeled compactness=0.736 Maximum likelihood:-Ln L=491.768 Modeled compactness at the center=0.166 Modeled compactness at the center=0.4 AIC=991.5361 Modeled compactness at the periphery=0.769 Modeled compactness at the periphery=0.933 r2=0.2558281 R/t (Currey & Alexander 1985) : 2.8788 R/t (Currey & Alexander 1985) : 2.4494 Observed compactness=0.684 CDI (Castanet et al. 2000) : 0.3474 CDI (Castanet et al. 2000) : 0.4083 Modeled compactness=0.67 From 342 ° to 347 ° Modeled compactness at the center=0.282 Maximum likelihood:-Ln L=194.7121 Angular analysis Modeled compactness at the periphery=0.962 AIC=397.4242 From 0 ° to 5 ° R/t (Currey & Alexander 1985) : 2.8446 r2=0.5312877 Maximum likelihood:-Ln L=290.6414 CDI (Castanet et al. 2000) : 0.3515 Observed compactness=0.462

219

AIC=589.2828 Modeled compactness at the periphery=0.964 AIC=677.3144 r2=0.5889634 R/t (Currey & Alexander 1985) : 2.3573 r2=0.5862562 Observed compactness=0.682 CDI (Castanet et al. 2000) : 0.4242 Observed compactness=0.77 Modeled compactness=0.71 From 30 ° to 35 ° Modeled compactness=0.768 Modeled compactness at the center=0.199 Maximum likelihood:-Ln L=351.7913 Modeled compactness at the center=0.18 Modeled compactness at the periphery=0.994 AIC=711.5825 Modeled compactness at the periphery=0.955 R/t (Currey & Alexander 1985) : 2.3684 r2=0.4515728 R/t (Currey & Alexander 1985) : 1.5612 CDI (Castanet et al. 2000) : 0.4222 Observed compactness=0.697 CDI (Castanet et al. 2000) : 0.6405 From 6 ° to 11 ° Modeled compactness=0.694 From 60 ° to 65 ° Maximum likelihood:-Ln L=271.5182 Modeled compactness at the center=0.329 Maximum likelihood:-Ln L=309.9672 AIC=551.0365 Modeled compactness at the periphery=0.962 AIC=627.9344 r2=0.6011457 R/t (Currey & Alexander 1985) : 2.7896 r2=0.5603214 Observed compactness=0.648 CDI (Castanet et al. 2000) : 0.3585 Observed compactness=0.798 Modeled compactness=0.65 From 36 ° to 41 ° Modeled compactness=0.801 Modeled compactness at the center=0.221 Maximum likelihood:-Ln L=435.3696 Modeled compactness at the center=0.174 Modeled compactness at the periphery=0.974 AIC=878.7391 Modeled compactness at the periphery=0.92 R/t (Currey & Alexander 1985) : 2.8107 r2=0.3127028 R/t (Currey & Alexander 1985) : 1.4558 CDI (Castanet et al. 2000) : 0.3558 Observed compactness=0.681 CDI (Castanet et al. 2000) : 0.6869 From 12 ° to 17 ° Modeled compactness=0.677 From 66 ° to 71 ° Maximum likelihood:-Ln L=343.9464 Modeled compactness at the center=0.339 Maximum likelihood:-Ln L=297.0038 AIC=695.8928 Modeled compactness at the periphery=0.928 AIC=602.0076 r2=0.6501501 R/t (Currey & Alexander 1985) : 2.9447 r2=0.5289236 Observed compactness=0.691 CDI (Castanet et al. 2000) : 0.3396 Observed compactness=0.779 Modeled compactness=0.687 From 42 ° to 47 ° Modeled compactness=0.783 Modeled compactness at the center=0.145 Maximum likelihood:-Ln L=351.0375 Modeled compactness at the center=0.228 Modeled compactness at the periphery=0.978 AIC=710.0751 Modeled compactness at the periphery=0.951 R/t (Currey & Alexander 1985) : 2.2991 r2=0.3316982 R/t (Currey & Alexander 1985) : 1.4574 CDI (Castanet et al. 2000) : 0.435 Observed compactness=0.739 CDI (Castanet et al. 2000) : 0.6862 From 18 ° to 23 ° Modeled compactness=0.736 From 72 ° to 77 ° Maximum likelihood:-Ln L=379.7088 Modeled compactness at the center=0.317 Maximum likelihood:-Ln L=370.478 AIC=767.4177 Modeled compactness at the periphery=0.918 AIC=748.9561 r2=0.4720502 R/t (Currey & Alexander 1985) : 1.8578 r2=0.4874541 Observed compactness=0.636 CDI (Castanet et al. 2000) : 0.5383 Observed compactness=0.737 Modeled compactness=0.633 From 48 ° to 53 ° Modeled compactness=0.737 Modeled compactness at the center=0.374 Maximum likelihood:-Ln L=312.7015 Modeled compactness at the center=0.328 Modeled compactness at the periphery=0.986 AIC=633.403 Modeled compactness at the periphery=0.821 R/t (Currey & Alexander 1985) : 4.0656 r2=0.5517284 R/t (Currey & Alexander 1985) : 1.6857 CDI (Castanet et al. 2000) : 0.246 Observed compactness=0.737 CDI (Castanet et al. 2000) : 0.5932 From 24 ° to 29 ° Modeled compactness=0.733 From 78 ° to 83 ° Maximum likelihood:-Ln L=347.5424 Modeled compactness at the center=0.159 Maximum likelihood:-Ln L=352.2104 AIC=703.0847 Modeled compactness at the periphery=0.971 AIC=712.4209 r2=0.557508 R/t (Currey & Alexander 1985) : 1.9618 r2=0.391438 Observed compactness=0.699 CDI (Castanet et al. 2000) : 0.5097 Observed compactness=0.818 Modeled compactness=0.707 From 54 ° to 59 ° Modeled compactness=0.817 Modeled compactness at the center=0.252 Maximum likelihood:-Ln L=334.6572 Modeled compactness at the center=0.353

220

Modeled compactness at the periphery=0.928 AIC=672.3372 Modeled compactness at the periphery=0.859 R/t (Currey & Alexander 1985) : 1.1747 r2=0.4659083 R/t (Currey & Alexander 1985) : 1.7288 CDI (Castanet et al. 2000) : 0.8513 Observed compactness=0.771 CDI (Castanet et al. 2000) : 0.5784 From 84 ° to 89 ° Modeled compactness=0.772 From 138 ° to 143 ° Maximum likelihood:-Ln L=278.8611 Modeled compactness at the center=0.35 Maximum likelihood:-Ln L=313.253 AIC=565.7221 Modeled compactness at the periphery=0.987 AIC=634.5059 r2=0.6388373 R/t (Currey & Alexander 1985) : 2.24 r2=0.5675631 Observed compactness=0.793 CDI (Castanet et al. 2000) : 0.4464 Observed compactness=0.704 Modeled compactness=0.795 From 114 ° to 119 ° Modeled compactness=0.701 Modeled compactness at the center=0.376 Maximum likelihood:-Ln L=322.4112 Modeled compactness at the center=0.187 Modeled compactness at the periphery=0.983 AIC=652.8224 Modeled compactness at the periphery=0.927 R/t (Currey & Alexander 1985) : 2.0739 r2=0.4919693 R/t (Currey & Alexander 1985) : 1.9373 CDI (Castanet et al. 2000) : 0.4822 Observed compactness=0.764 CDI (Castanet et al. 2000) : 0.5162 From 90 ° to 95 ° Modeled compactness=0.764 From 144 ° to 149 ° Maximum likelihood:-Ln L=286.0774 Modeled compactness at the center=0.196 Maximum likelihood:-Ln L=262.1996 AIC=580.1547 Modeled compactness at the periphery=0.933 AIC=532.3991 r2=0.6293671 R/t (Currey & Alexander 1985) : 1.5138 r2=0.2497238 Observed compactness=0.665 CDI (Castanet et al. 2000) : 0.6606 Observed compactness=0.701 Modeled compactness=0.683 From 120 ° to 125 ° Modeled compactness=0.704 Modeled compactness at the center=0.211 Maximum likelihood:-Ln L=339.7931 Modeled compactness at the center=0.602 Modeled compactness at the periphery=0.978 AIC=687.5862 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 2.5074 r2=0.4263514 R/t (Currey & Alexander 1985) : 7.1169 CDI (Castanet et al. 2000) : 0.3988 Observed compactness=0.783 CDI (Castanet et al. 2000) : 0.1405 From 96 ° to 101 ° Modeled compactness=0.786 From 150 ° to 155 ° Maximum likelihood:-Ln L=256.6036 Modeled compactness at the center=0.45 Maximum likelihood:-Ln L=250.1291 AIC=521.2072 Modeled compactness at the periphery=0.987 AIC=508.2582 r2=0.6272731 R/t (Currey & Alexander 1985) : 2.4497 r2=0.4742818 Observed compactness=0.68 CDI (Castanet et al. 2000) : 0.4082 Observed compactness=0.607 Modeled compactness=0.685 From 126 ° to 131 ° Modeled compactness=0.613 Modeled compactness at the center=0.245 Maximum likelihood:-Ln L=243.043 Modeled compactness at the center=0.307 Modeled compactness at the periphery=0.981 AIC=494.0861 Modeled compactness at the periphery=0.86 R/t (Currey & Alexander 1985) : 2.609 r2=0.3625667 R/t (Currey & Alexander 1985) : 3.5584 CDI (Castanet et al. 2000) : 0.3833 Observed compactness=0.707 CDI (Castanet et al. 2000) : 0.281 From 102 ° to 107 ° Modeled compactness=0.705 From 156 ° to 161 ° Maximum likelihood:-Ln L=311.9335 Modeled compactness at the center=0.587 Maximum likelihood:-Ln L=321.6418 AIC=631.8671 Modeled compactness at the periphery=1. AIC=651.2836 r2=0.6596577 R/t (Currey & Alexander 1985) : 6.2966 r2=0.4474721 Observed compactness=0.708 CDI (Castanet et al. 2000) : 0.1588 Observed compactness=0.693 Modeled compactness=0.722 From 132 ° to 137 ° Modeled compactness=0.692 Modeled compactness at the center=0.168 Maximum likelihood:-Ln L=359.624 Modeled compactness at the center=0.186 Modeled compactness at the periphery=0.937 AIC=727.248 Modeled compactness at the periphery=0.911 R/t (Currey & Alexander 1985) : 2.0044 r2=0.5504249 R/t (Currey & Alexander 1985) : 1.8739 CDI (Castanet et al. 2000) : 0.4989 Observed compactness=0.66 CDI (Castanet et al. 2000) : 0.5336 From 108 ° to 113 ° Modeled compactness=0.665 From 162 ° to 167 ° Maximum likelihood:-Ln L=332.1686 Modeled compactness at the center=0.215 Maximum likelihood:-Ln L=246.1299

221

AIC=500.2598 Modeled compactness at the periphery=0.991 AIC=497.4332 r2=0.415536 R/t (Currey & Alexander 1985) : 3.9294 r2=0.1344952 Observed compactness=0.645 CDI (Castanet et al. 2000) : 0.2545 Observed compactness=0.543 Modeled compactness=0.644 From 192 ° to 197 ° Modeled compactness=0.54 Modeled compactness at the center=0.273 Maximum likelihood:-Ln L=351.0436 Modeled compactness at the center=0.505 Modeled compactness at the periphery=0.8 AIC=710.0872 Modeled compactness at the periphery=0.996 R/t (Currey & Alexander 1985) : 1.5512 r2=0.2519763 R/t (Currey & Alexander 1985) : 27.7133 CDI (Castanet et al. 2000) : 0.6447 Observed compactness=0.786 CDI (Castanet et al. 2000) : 0.0361 From 168 ° to 173 ° Modeled compactness=0.789 From 222 ° to 227 ° Maximum likelihood:-Ln L=164.2717 Modeled compactness at the center=0.692 Maximum likelihood:-Ln L=365.9369 AIC=336.5434 Modeled compactness at the periphery=1. AIC=739.8737 r2=0.617439 R/t (Currey & Alexander 1985) : 5.637 r2=0.1635616 Observed compactness=0.618 CDI (Castanet et al. 2000) : 0.1774 Observed compactness=0.672 Modeled compactness=0.614 From 198 ° to 203 ° Modeled compactness=0.67 Modeled compactness at the center=0.097 Maximum likelihood:-Ln L=374.1083 Modeled compactness at the center=0.444 Modeled compactness at the periphery=0.875 AIC=756.2167 Modeled compactness at the periphery=0.715 R/t (Currey & Alexander 1985) : 2.1572 r2=0.1890909 R/t (Currey & Alexander 1985) : 1.6809 CDI (Castanet et al. 2000) : 0.4636 Observed compactness=0.779 CDI (Castanet et al. 2000) : 0.5949 From 174 ° to 179 ° Modeled compactness=0.782 From 228 ° to 233 ° Maximum likelihood:-Ln L=151.7753 Modeled compactness at the center=0.7 Maximum likelihood:-Ln L=136.9606 AIC=311.5506 Modeled compactness at the periphery=1. AIC=281.9211 r2=0.3507972 R/t (Currey & Alexander 1985) : 6.5733 r2=0.4980223 Observed compactness=0.529 CDI (Castanet et al. 2000) : 0.1521 Observed compactness=0.41 Modeled compactness=0.517 From 204 ° to 209 ° Modeled compactness=0.402 Modeled compactness at the center=0.033 Maximum likelihood:-Ln L=254.8957 Modeled compactness at the center=0.26 Modeled compactness at the periphery=0.635 AIC=517.7915 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 1.599 r2=0.2102492 R/t (Currey & Alexander 1985) : 9.7369 CDI (Castanet et al. 2000) : 0.6254 Observed compactness=0.762 CDI (Castanet et al. 2000) : 0.1027 From 180 ° to 185 ° Modeled compactness=0.764 From 234 ° to 239 ° Maximum likelihood:-Ln L=216.9165 Modeled compactness at the center=0.702 Maximum likelihood:-Ln L=204.5007 AIC=441.833 Modeled compactness at the periphery=1. AIC=417.0015 r2=0.3511049 R/t (Currey & Alexander 1985) : 9.0199 r2=0.1125672 Observed compactness=0.837 CDI (Castanet et al. 2000) : 0.1109 Observed compactness=0.62 Modeled compactness=0.831 From 210 ° to 215 ° Modeled compactness=0.616 Modeled compactness at the center=0.595 Maximum likelihood:-Ln L=382.4985 Modeled compactness at the center=0.566 Modeled compactness at the periphery=0.985 AIC=772.9969 Modeled compactness at the periphery=1. R/t (Currey & Alexander 1985) : 2.578 r2=0.1716028 R/t (Currey & Alexander 1985) : 15.884 CDI (Castanet et al. 2000) : 0.3879 Observed compactness=0.681 CDI (Castanet et al. 2000) : 0.063 From 186 ° to 191 ° Modeled compactness=0.677 From 240 ° to 245 ° Maximum likelihood:-Ln L=324.1383 Modeled compactness at the center=0.585 Maximum likelihood:-Ln L=153.5051 AIC=656.2766 Modeled compactness at the periphery=1. AIC=315.0101 r2=0.260347 R/t (Currey & Alexander 1985) : 8.2693 r2=0.0737018 Observed compactness=0.81 CDI (Castanet et al. 2000) : 0.1209 Observed compactness=0.698 Modeled compactness=0.808 From 216 ° to 221 ° Modeled compactness=0.693 Modeled compactness at the center=0.666 Maximum likelihood:-Ln L=244.7166 Modeled compactness at the center=0.667

222

Modeled compactness at the periphery=1. AIC=851.862 Modeled compactness at the periphery=0.893 R/t (Currey & Alexander 1985) : 25.3856 r2=0.3665841 R/t (Currey & Alexander 1985) : 1.5749 CDI (Castanet et al. 2000) : 0.0394 Observed compactness=0.748 CDI (Castanet et al. 2000) : 0.635 From 246 ° to 251 ° Modeled compactness=0.749 From 300 ° to 305 ° Maximum likelihood:-Ln L=128.2353 Modeled compactness at the center=0.409 Maximum likelihood:-Ln L=461.0331 AIC=264.4706 Modeled compactness at the periphery=0.956 AIC=930.0662 r2=0.2319461 R/t (Currey & Alexander 1985) : 2.4874 r2=0.4697965 Observed compactness=0.702 CDI (Castanet et al. 2000) : 0.402 Observed compactness=0.656 Modeled compactness=0.698 From 276 ° to 281 ° Modeled compactness=0.651 Modeled compactness at the center=0.528 Maximum likelihood:-Ln L=263.532 Modeled compactness at the center=0.18 Modeled compactness at the periphery=0.952 AIC=535.064 Modeled compactness at the periphery=0.933 R/t (Currey & Alexander 1985) : 4.7932 r2=0.6983789 R/t (Currey & Alexander 1985) : 2.4722 CDI (Castanet et al. 2000) : 0.2086 Observed compactness=0.807 CDI (Castanet et al. 2000) : 0.4045 From 252 ° to 257 ° Modeled compactness=0.805 From 306 ° to 311 ° Maximum likelihood:-Ln L=128.8759 Modeled compactness at the center=0.004 Maximum likelihood:-Ln L=350.6226 AIC=265.7518 Modeled compactness at the periphery=0.84 AIC=709.2452 r2=0.0391651 R/t (Currey & Alexander 1985) : 1.2372 r2=0.253262 Observed compactness=0.611 CDI (Castanet et al. 2000) : 0.8083 Observed compactness=0.74 Modeled compactness=0.611 From 282 ° to 287 ° Modeled compactness=0.735 Modeled compactness at the center=0.602 Maximum likelihood:-Ln L=296.9389 Modeled compactness at the center=0.605 Modeled compactness at the periphery=0.615 AIC=601.8777 Modeled compactness at the periphery=0.995 R/t (Currey & Alexander 1985) : 0.8848 r2=0.17427 R/t (Currey & Alexander 1985) : 5.3187 CDI (Castanet et al. 2000) : 1.1301 Observed compactness=0.766 CDI (Castanet et al. 2000) : 0.188 From 258 ° to 263 ° Modeled compactness=0.768 From 312 ° to 317 ° Maximum likelihood:-Ln L=105.4834 Modeled compactness at the center=0.683 Maximum likelihood:-Ln L=342.7057 AIC=218.9669 Modeled compactness at the periphery=0.959 AIC=693.4114 r2=0.2264045 R/t (Currey & Alexander 1985) : 5.8679 r2=0.5129865 Observed compactness=0.514 CDI (Castanet et al. 2000) : 0.1704 Observed compactness=0.729 Modeled compactness=0.508 From 288 ° to 293 ° Modeled compactness=0.719 Modeled compactness at the center=0.393 Maximum likelihood:-Ln L=265.7792 Modeled compactness at the center=0.358 Modeled compactness at the periphery=0.901 AIC=539.5584 Modeled compactness at the periphery=0.937 R/t (Currey & Alexander 1985) : 10.3595 r2=0.7008185 R/t (Currey & Alexander 1985) : 2.4996 CDI (Castanet et al. 2000) : 0.0965 Observed compactness=0.788 CDI (Castanet et al. 2000) : 0.4001 From 264 ° to 269 ° Modeled compactness=0.787 From 318 ° to 323 ° Maximum likelihood:-Ln L=119.5569 Modeled compactness at the center=0. Maximum likelihood:-Ln L=398.483 AIC=247.1137 Modeled compactness at the periphery=0.811 AIC=804.966 r2=0.5033083 R/t (Currey & Alexander 1985) : 1.2035 r2=0.4772329 Observed compactness=0.507 CDI (Castanet et al. 2000) : 0.8309 Observed compactness=0.776 Modeled compactness=0.517 From 294 ° to 299 ° Modeled compactness=0.773 Modeled compactness at the center=0. Maximum likelihood:-Ln L=363.5078 Modeled compactness at the center=0.31 Modeled compactness at the periphery=0.56 AIC=735.0157 Modeled compactness at the periphery=0.923 R/t (Currey & Alexander 1985) : 1.3731 r2=0.4113331 R/t (Currey & Alexander 1985) : 1.3335 CDI (Castanet et al. 2000) : 0.7283 Observed compactness=0.715 CDI (Castanet et al. 2000) : 0.7499 From 270 ° to 275 ° Modeled compactness=0.712 From 324 ° to 329 ° Maximum likelihood:-Ln L=421.931 Modeled compactness at the center=0.245 Maximum likelihood:-Ln L=340.243

223

AIC=688.4861 Modeled compactness at the periphery=0.976 Modeled compactness at the center=0. r2=0.4285414 R/t (Currey & Alexander 1985) : 2.0748 Modeled compactness at the periphery=0. Observed compactness=0.814 CDI (Castanet et al. 2000) : 0.482 R/t (Currey & Alexander 1985) : 1. Modeled compactness=0.814 From 354 ° to 359 ° CDI (Castanet et al. 2000) : 1. Modeled compactness at the center=0.429 Maximum likelihood:-Ln L=397.3978 From 12 ° to 17 ° Modeled compactness at the periphery=0.93 AIC=802.7957 Maximum likelihood:-Ln L=6066.291 R/t (Currey & Alexander 1985) : 1.1204 r2=0.5144731 AIC=12140.58 CDI (Castanet et al. 2000) : 0.8925 Observed compactness=0.778 r2=-1.#IND From 330 ° to 335 ° Modeled compactness=0.77 Observed compactness=0.531 Maximum likelihood:-Ln L=405.725 Modeled compactness at the center=0.266 Modeled compactness=0. AIC=819.4501 Modeled compactness at the periphery=0.953 Modeled compactness at the center=0. r2=0.2260041 R/t (Currey & Alexander 1985) : 1.7069 Modeled compactness at the periphery=0. Observed compactness=0.828 CDI (Castanet et al. 2000) : 0.5859 R/t (Currey & Alexander 1985) : 1. Modeled compactness=0.829 CDI (Castanet et al. 2000) : 1. Modeled compactness at the center=0.655 Cat From 18 ° to 23 ° Modeled compactness at the periphery=0.907 Maximum likelihood:-Ln L=5289.51

R/t (Currey & Alexander 1985) : 1.6654 AIC=10587.02 Global analysis CDI (Castanet et al. 2000) : 0.6005 r2=-1.#IND Maximum likelihood:-Ln L=295.1159 From 336 ° to 341 ° Observed compactness=0.53 AIC=598.2318 Maximum likelihood:-Ln L=388.2525 Modeled compactness=0. r2=0.99919 AIC=784.5049 Modeled compactness at the center=0. Observed compactness=0.541 r2=0.2961356 Modeled compactness at the periphery=0. Modeled compactness=0.541 Observed compactness=0.796 R/t (Currey & Alexander 1985) : 1. Modeled compactness at the center=0. Modeled compactness=0.797 CDI (Castanet et al. 2000) : 1. Modeled compactness at the periphery=0.976 Modeled compactness at the center=0.562 From 24 ° to 29 ° R/t (Currey & Alexander 1985) : 2.9688 Modeled compactness at the periphery=0.903 Maximum likelihood:-Ln L=5727.417 CDI (Castanet et al. 2000) : 0.3368 R/t (Currey & Alexander 1985) : 1.7849 AIC=11462.83

CDI (Castanet et al. 2000) : 0.5603 r2=-1.#IND Angular analysis From 342 ° to 347 ° Observed compactness=0.542 From 0 ° to 5 ° Maximum likelihood:-Ln L=352.7352 Modeled compactness=0. Maximum likelihood:-Ln L=5199.58 AIC=713.4703 Modeled compactness at the center=0. AIC=10407.16 r2=0.5632666 Modeled compactness at the periphery=0. r2=-1.#IND Observed compactness=0.842 R/t (Currey & Alexander 1985) : 1. Observed compactness=0.546 Modeled compactness=0.84 CDI (Castanet et al. 2000) : 1. Modeled compactness=0. Modeled compactness at the center=0.41 From 30 ° to 35 ° Modeled compactness at the center=0. Modeled compactness at the periphery=0.954 Maximum likelihood:-Ln L=5526.903 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1.1309 AIC=11061.81 R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 0.8842 r2=-1.#IND CDI (Castanet et al. 2000) : 1. From 348 ° to 353 ° Observed compactness=0.546 From 6 ° to 11 ° Maximum likelihood:-Ln L=296.8634 Modeled compactness=0. Maximum likelihood:-Ln L=5664.548 AIC=601.7268 Modeled compactness at the center=0. AIC=11337.1 r2=0.5342621 Modeled compactness at the periphery=0. r2=-1.#IND Observed compactness=0.811 R/t (Currey & Alexander 1985) : 1. Observed compactness=0.531 Modeled compactness=0.814 CDI (Castanet et al. 2000) : 1. Modeled compactness=0. Modeled compactness at the center=0.463 From 36 ° to 41 °

224

Maximum likelihood:-Ln L=5904.343 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5697.999 AIC=11816.69 Modeled compactness at the periphery=0. AIC=11404 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.547 CDI (Castanet et al. 2000) : 1. Observed compactness=0.526 Modeled compactness=0. From 66 ° to 71 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4502.283 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=9012.566 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.535 CDI (Castanet et al. 2000) : 1. From 42 ° to 47 ° Modeled compactness=0. From 96 ° to 101 ° Maximum likelihood:-Ln L=5384.752 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5666.432 AIC=10777.5 Modeled compactness at the periphery=0. AIC=11340.86 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.535 CDI (Castanet et al. 2000) : 1. Observed compactness=0.556 Modeled compactness=0. From 72 ° to 77 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5162.698 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=10333.4 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.565 CDI (Castanet et al. 2000) : 1. From 48 ° to 53 ° Modeled compactness=0. From 102 ° to 107 ° Maximum likelihood:-Ln L=5869.79 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5079.198 AIC=11747.58 Modeled compactness at the periphery=0. AIC=10166.4 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.547 CDI (Castanet et al. 2000) : 1. Observed compactness=0.538 Modeled compactness=0. From 78 ° to 83 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4325.42 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=8658.84 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.551 CDI (Castanet et al. 2000) : 1. From 54 ° to 59 ° Modeled compactness=0. From 108 ° to 113 ° Maximum likelihood:-Ln L=5226.842 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4793.521 AIC=10461.68 Modeled compactness at the periphery=0. AIC=9595.041 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.558 CDI (Castanet et al. 2000) : 1. Observed compactness=0.534 Modeled compactness=0. From 84 ° to 89 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=3585.817 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=7179.635 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.47 CDI (Castanet et al. 2000) : 1. From 60 ° to 65 ° Modeled compactness=0. From 114 ° to 119 ° Maximum likelihood:-Ln L=5141.817 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4359.896 AIC=10291.63 Modeled compactness at the periphery=0. AIC=8727.792 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.551 CDI (Castanet et al. 2000) : 1. Observed compactness=0.498 Modeled compactness=0. From 90 ° to 95 ° Modeled compactness=0.

225

Modeled compactness at the center=0. Maximum likelihood:-Ln L=4453.68 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=8915.359 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.524 CDI (Castanet et al. 2000) : 1. From 120 ° to 125 ° Modeled compactness=0. From 174 ° to 179 ° Maximum likelihood:-Ln L=4457.498 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5261.002 AIC=8922.995 Modeled compactness at the periphery=0. AIC=10530 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.529 CDI (Castanet et al. 2000) : 1. Observed compactness=0.591 Modeled compactness=0. From 150 ° to 155 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4209.7 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=8427.399 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.497 CDI (Castanet et al. 2000) : 1. From 126 ° to 131 ° Modeled compactness=0. From 180 ° to 185 ° Maximum likelihood:-Ln L=4741.005 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4474.407 AIC=9490.011 Modeled compactness at the periphery=0. AIC=8956.814 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.537 CDI (Castanet et al. 2000) : 1. Observed compactness=0.502 Modeled compactness=0. From 156 ° to 161 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4299.621 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=8607.241 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.517 CDI (Castanet et al. 2000) : 1. From 132 ° to 137 ° Modeled compactness=0. From 186 ° to 191 ° Maximum likelihood:-Ln L=4623.055 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4577.971 AIC=9254.109 Modeled compactness at the periphery=0. AIC=9163.942 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.51 CDI (Castanet et al. 2000) : 1. Observed compactness=0.531 Modeled compactness=0. From 162 ° to 167 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5085.891 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=10179.78 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.55 CDI (Castanet et al. 2000) : 1. From 138 ° to 143 ° Modeled compactness=0. From 192 ° to 197 ° Maximum likelihood:-Ln L=4515.145 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5921.756 AIC=9038.29 Modeled compactness at the periphery=0. AIC=11851.51 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.528 CDI (Castanet et al. 2000) : 1. Observed compactness=0.552 Modeled compactness=0. From 168 ° to 173 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5440.421 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=10888.84 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.561 CDI (Castanet et al. 2000) : 1. From 144 ° to 149 ° Modeled compactness=0. From 198 ° to 203 °

226

Maximum likelihood:-Ln L=5622.805 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5492.052 AIC=11253.61 Modeled compactness at the periphery=0. AIC=10992.1 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.524 CDI (Castanet et al. 2000) : 1. Observed compactness=0.559 Modeled compactness=0. From 228 ° to 233 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=6094.122 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=12196.24 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.573 CDI (Castanet et al. 2000) : 1. From 204 ° to 209 ° Modeled compactness=0. From 258 ° to 263 ° Maximum likelihood:-Ln L=5408.724 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5221.924 AIC=10825.45 Modeled compactness at the periphery=0. AIC=10451.85 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.544 CDI (Castanet et al. 2000) : 1. Observed compactness=0.519 Modeled compactness=0. From 234 ° to 239 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5754.03 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=11516.06 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.576 CDI (Castanet et al. 2000) : 1. From 210 ° to 215 ° Modeled compactness=0. From 264 ° to 269 ° Maximum likelihood:-Ln L=5807.801 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4337.378 AIC=11623.6 Modeled compactness at the periphery=0. AIC=8682.756 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.533 CDI (Castanet et al. 2000) : 1. Observed compactness=0.494 Modeled compactness=0. From 240 ° to 245 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4983.674 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=9975.348 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.525 CDI (Castanet et al. 2000) : 1. From 216 ° to 221 ° Modeled compactness=0. From 270 ° to 275 ° Maximum likelihood:-Ln L=5212.083 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5688.663 AIC=10432.17 Modeled compactness at the periphery=0. AIC=11385.33 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.505 CDI (Castanet et al. 2000) : 1. Observed compactness=0.538 Modeled compactness=0. From 246 ° to 251 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5688.899 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=11385.8 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.565 CDI (Castanet et al. 2000) : 1. From 222 ° to 227 ° Modeled compactness=0. From 276 ° to 281 ° Maximum likelihood:-Ln L=5517.435 Modeled compactness at the center=0. Maximum likelihood:-Ln L=5066.91 AIC=11042.87 Modeled compactness at the periphery=0. AIC=10141.82 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.553 CDI (Castanet et al. 2000) : 1. Observed compactness=0.525 Modeled compactness=0. From 252 ° to 257 ° Modeled compactness=0.

227

Modeled compactness at the center=0. Maximum likelihood:-Ln L=5507.29 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=11022.58 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.583 CDI (Castanet et al. 2000) : 1. From 282 ° to 287 ° Modeled compactness=0. From 336 ° to 341 ° Maximum likelihood:-Ln L=5363.57 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4604.442 AIC=10735.14 Modeled compactness at the periphery=0. AIC=9216.884 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.543 CDI (Castanet et al. 2000) : 1. Observed compactness=0.534 Modeled compactness=0. From 312 ° to 317 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4933.509 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=9875.018 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.546 CDI (Castanet et al. 2000) : 1. From 288 ° to 293 ° Modeled compactness=0. From 342 ° to 347 ° Maximum likelihood:-Ln L=5184.265 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4723.877 AIC=10376.53 Modeled compactness at the periphery=0. AIC=9455.755 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.547 CDI (Castanet et al. 2000) : 1. Observed compactness=0.587 Modeled compactness=0. From 318 ° to 323 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=5063.453 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=10134.91 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.557 CDI (Castanet et al. 2000) : 1. From 294 ° to 299 ° Modeled compactness=0. From 348 ° to 353 ° Maximum likelihood:-Ln L=5076.734 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4802.61 AIC=10161.47 Modeled compactness at the periphery=0. AIC=9613.22 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.542 CDI (Castanet et al. 2000) : 1. Observed compactness=0.585 Modeled compactness=0. From 324 ° to 329 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4637.1 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=9282.199 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.56 CDI (Castanet et al. 2000) : 1. From 300 ° to 305 ° Modeled compactness=0. From 354 ° to 359 ° Maximum likelihood:-Ln L=5097.565 Modeled compactness at the center=0. Maximum likelihood:-Ln L=4721.75 AIC=10203.13 Modeled compactness at the periphery=0. AIC=9451.499 r2=-1.#IND R/t (Currey & Alexander 1985) : 1. r2=-1.#IND Observed compactness=0.543 CDI (Castanet et al. 2000) : 1. Observed compactness=0.584 Modeled compactness=0. From 330 ° to 335 ° Modeled compactness=0. Modeled compactness at the center=0. Maximum likelihood:-Ln L=4396.447 Modeled compactness at the center=0. Modeled compactness at the periphery=0. AIC=8800.895 Modeled compactness at the periphery=0. R/t (Currey & Alexander 1985) : 1. r2=-1.#IND R/t (Currey & Alexander 1985) : 1. CDI (Castanet et al. 2000) : 1. Observed compactness=0.544 CDI (Castanet et al. 2000) : 1. From 306 ° to 311 ° Modeled compactness=0.

228

Appendix C NSM.005.GF.045.044

229

230

NSM.005.GF.045.o47-D

231

232

NSM.007.GF.004.630

233

234

NSM.005.GF.045.048-A

235

236

RM.20.6711

237

238

Platypus (MVZ M35370-70863)

239

240

Cat

241

Appendix D Fossil bone segmentation

1. Load the dataset

242

2. Open the image processing toolbox

243

3. Apply a median filter, a commonly used kernel size is 3

244

4. Apply a mean shift filter, making sure the kernel size is 2 larger than what was used in the median filter

245

5. Threshold the background, then use the background to overwrite the background of the median- mean-shift dataset. Overwrite using the value of the bone density (can be determined using the probe tool). Make sure the background ROI is selected in the data properties box, and the correct dataset is selected to be overwritten in the dataset operations toolbox.

246

Overwritten background

247

6. In the image processing toolbox apply an arithmetic filter converting the dataset to a 32-bit float file (necessary for later Distance of Gaussian (DoG) filter)

248

7. Apply a polynomial function of the third degree (to aid in correction of beam hardening)

249

8. Apply a DoG filter with first sigma ranging between 0.5 and 1.0, second sigma between 2.0 and 4.0. Make sure the normalize box is unchecked. Compute previews to determine which combinations of these filters and their parameters are best before applying them to the whole dataset.

250

9. This is the dataset after application of the filters.

251

10. Define a range using the heavily filtered (median-mean-shift-arithmetic-polynomial-DoG) that accurately captures the trabecular structure. Don’t worry about the ROI including parts of the background, it can be removed later.

252

11. Threshold an ROI using the less filtered (median-mean-shift) dataset, which accurately covers the cortical bone.

253

12. Merge (union; Boolean Operations) the ROIs from steps 10 and 11, and remove the background (A-B; Boolean Operations)

254

255

13. Segment the Cortical and Trabecular bone of the merged ROI (Tools; Bone Analysis)

256

14. Using the Kohler segmentation you can manually adjust the threshold to get results that most accurately reflect the cortical and trabecular bone.

257

15. Four output ROIs from the segmentation, note the difference between bone and area ROIs

258

259

16. Right click the trabecular bone ROI to run a Multi-ROI analysis

260

17. select the tools button to run a volume analysis, and select the largest volume to export as an ROI, this will denoise the trabecular dataset (purple non-trabecular noise is removed)

261

18. Create a box that is your whole volume you will be sampling

262

19. Using the ROI from the Multi-ROI analysis compute the mappings with the parameters required (outlined in the text). Spacing and radius of influence make up the smaller volumes of interest sampled within the whole bone volume

263

20. Through a union of the Trabecular area and Cortical area create an infill (whole bone) ROI

264

21. For measurement data make sure the bone segmentataion is the infull ROI and the Cortical and Trabecular BONE ROIs are used. Note that in the version of dragonfly I was using the average trabecular separation would not work, BV and TV were reversed and single whole bone measurement of anisotropy would not provide any kind of meaningful data.

265

Appendix E Supplemental trabecular methods (extant)

1. Define a range

266

267

2. Export range to ROI

268

3. Run a multi ROI analysis to separate elements, or clean up a single element (remove high intensity inclusions that are not connected trabeculae

269

270

4. Automatically segment cortical bone ant trabecular bone, using the segment tab in the Bone analysis window (manually add areas of cortical bone that were incorrectly assigned, then subtract the cortical bone ROI from the trabecular bone ROI to get more accurate trabecular and

271

cortical bone ROIs)

5. Using A filled in ROI of the bone of interest (Can be done by using the “fill inner area” in the X Y and Z planes from the isolated humerus, or combining trabecular and cortical areas from the segmentation), the Cortical bone ROI and trabecular bone ROI calculate average trabecular thickness. At this point most measurements can be computed.

272

273

6. Set up boxes that enclose your area of interest. I used a bow that enclosed the whole bone, and one that took a slice in the middle of the bone alligned with the orientation of the proximal articulation.

274

7. Using the surface normals algorithm on your best segmentation of trabecular bone within the box enclosing your area of interest assign sampling parameters. Here I used 500um which was t least 3x as wide as the average trabecular thickness. I used identicle samping and radius of influence in an attempt to reduce noise but maintain some overlap to detect subtle changes.

275

8. You get 2 output datasets from this, a vector map and scalar map of anisotropy, below is the vector map which can be visually changed using LUTs to make reading them easier. Here Jet was used. The values of anisotropy can be thresholded as well so visually only specific values are seen if you choose.

276

9. Invert the trabecular area, which was one of the 4 outputs from the cortical bone segmentation (this is because the sampling picks up the trabecular cortical “cut” surface and it gets mapped as highly anisotropic, here we will remove the noise from those edges)

277

10. Convert the scalar anisotropy map to jet –for visual ease

278

11. Convert the map to nearest neighbour interpolation

279

12. Overwrite the inverted trabecular area on the scalar anisotropy dataset with a value of 0

280

281

13. Convert scalar map back to desired interpolation

14. This will give you an accurate frequency of anisotropy histogram

282

Appendix F PCA femora only with angular bone profiler variables Table S.0.1 PCA summary statistics

PC Eigenvalue % variance 1 3.75181 46.898 2 1.72683 21.585 3 1.17943 14.743 4 0.833091 10.414 5 0.358108 4.4764 6 0.098205 1.2276 7 0.047553 0.59441 8 0.004969 0.062106

Table S.0.2 PCA loadings

PC 1 PC 2 PC 3 PC 4 PC 5 PC 6 PC 7 PC 8 BV/TV 0.28702 -0.46192 0.13951 0.55678 0.2766 0.34016 -0.19263 -0.38139 Ct.Ar/Tt.Ar 0.49127 0.082318 0.00738 0.044453 0.46653 0.072354 0.030048 0.7253 Ani.M 0.29058 0.53465 0.28803 -0.14826 0.42909 -0.22113 0.17099 -0.51236 Rmin -0.28664 0.20915 -0.44344 0.65405 0.19422 -0.29257 0.35462 0.024407 Rmax 0.47915 0.041008 -0.15465 0.024464 -0.46391 0.31861 0.64769 -0.08934 RS -0.44708 -0.16936 -0.03703 -0.31722 0.47641 0.50322 0.43377 -0.03275 RP -0.20423 0.62973 0.19775 0.2824 -0.15959 0.59006 -0.24282 0.10139 PD length -0.19622 -0.15276 0.7978 0.23931 -0.1279 -0.20407 0.37466 0.21456 (cm)

Recall that turtle midshaft profile are anomalous to other amniote midshaft patterns, therefore Elseya the only definitive aquatic femora may not be a good comparison for this PCA. It is however the only definitive aquatic femur I had access to.

283

284

285

286

287

Appendix G PCA of all elements, Felis removed

Table S.1 PCA no Felis summary

PC Eigenvalue % variance 1 2.18501 54.625 2 1.31897 32.974 3 0.387985 9.6996 4 0.108037 2.7009

Table S.2 PCA no Felis loadings

PC 1 PC 2 PC 3 PC 4 BV/TV 0.54032 0.41836 -0.51291 -0.51956 CT.Ar/Tt.Ar 0.64645 -0.11758 -0.16092 0.73646 Ani.M 0.53369 -0.39203 0.63896 -0.39144 PD (cm) 0.072984 0.81084 0.55023 0.18562

288

289